diff --git a/.capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml b/.capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml
index 88b9159..bdbb6f6 100644
--- a/.capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml
+++ b/.capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml
@@ -18,7 +18,10 @@ guidance:
   - Periods are fixed 24-hour windows, not daylight-saving-aware local calendar
     days.
   - Native and daily views share source observations and are correlated.
-  - This recipe supports schema 1.0 and does not support reset_points.
+  - Legacy schema 1.0 requires complete input and does not support resets.
+    Schema 1.1 labels sparse native/completed-day views and preserves reset
+    boundaries; daily sums require declared additive volume and fully observed
+    children.
 planning:
   story_map:
     backbone: Monitor while periods are still open
@@ -41,6 +44,8 @@ agent:
     - analysis/expectations/model-seasonal-level-and-growth
   implementation:
     references:
+      - src/anomalyzer/sparse_views.py
+      - tests/test_sparse.py
       - src/anomalyzer/recipes.py
       - tests/test_analysis.py
       - tests/test_capability_behaviors.py
diff --git a/.capabilities/data/inputs/accept-analysis-requests.capability.yaml b/.capabilities/data/inputs/accept-analysis-requests.capability.yaml
index 01a6e22..204bf4d 100644
--- a/.capabilities/data/inputs/accept-analysis-requests.capability.yaml
+++ b/.capabilities/data/inputs/accept-analysis-requests.capability.yaml
@@ -14,8 +14,9 @@ acceptance:
   - Published schema 1.0 and 1.1 request and result schemas match the public
     Python contracts.
 guidance:
-  - Null observations represent missing values, never zero. Limits apply to the
-    request as a whole.
+  - Null observations remain unknown unless a schema-1.1 additive-volume policy
+    explicitly authorizes confirmed empty collection as zero. Limits apply to
+    the request as a whole.
   - Describe current supported behavior; historical plans do not add accepted
     inputs.
 planning:
diff --git a/.capabilities/data/quality/analyze-sparse-observations.capability.yaml b/.capabilities/data/quality/analyze-sparse-observations.capability.yaml
new file mode 100644
index 0000000..e88c986
--- /dev/null
+++ b/.capabilities/data/quality/analyze-sparse-observations.capability.yaml
@@ -0,0 +1,321 @@
+title: Assess sparse observations without inventing consumption
+status: implemented
+summary: Schema-1.1 callers declare coverage and missing-data semantics per
+  source, keeping unknown collection distinct from confirmed empty additive
+  volume.
+intent: Retain useful rule and supported statistical evidence for sparse
+  histories while preserving time, provenance, incomplete periods and evidence
+  limits.
+acceptance:
+  - Bounded calendar preparation preserves phase, original values and coverage;
+    only an explicit additive-volume empty policy authorizes zero conversion.
+  - Frozen eligible training and calibration support point references across
+    gaps, while pattern and change comparisons respect gaps and explicit resets.
+  - Event-index profiles preserve real event timestamps and reject calendar
+    seasonality; unavailable relationship inputs and zero denominators stay
+    unavailable.
+  - As-of replay respects supplied arrival/revision timestamps and sparse
+    multi-resolution views retain useful siblings without inventing complete
+    totals.
+guidance:
+  - Default missingness is unknown. A modeled expectation is never a measurement.
+  - Caller business severity, incident state and notifications remain outside
+    the engine.
+planning:
+  story_map:
+    backbone: Monitor while periods are still open
+    step: Assess sparse source data
+    release: Incomplete-period monitoring
+    order: 30
+#
+#
+agent:
+  # -----------------------------
+  # machine managed agent metadata
+  # review all capabilities and save evidence: capabilitykit review
+  # review this capability and save evidence: capabilitykit review data/quality/analyze-sparse-observations
+  # run deterministic review only: capabilitykit review data/quality/analyze-sparse-observations --deterministic-only
+  # ask an agent and save review evidence: capabilitykit review data/quality/analyze-sparse-observations --agent codex --arg exec --handoff stdin
+  # validate saved agent output without writing: capabilitykit review-result data/quality/analyze-sparse-observations --input review.json
+  # -----------------------------
+  depends_on:
+    - data/inputs/accept-analysis-requests
+    - analysis/expectations/reset-declared-regimes
+    - analysis/detection/identify-residual-patterns
+  implementation:
+    references:
+      - src/anomalyzer/contracts.py
+      - src/anomalyzer/sparse.py
+      - src/anomalyzer/sparse_views.py
+      - src/anomalyzer/orchestrator.py
+      - src/anomalyzer/relationships.py
+      - src/anomalyzer/early.py
+      - tests/test_sparse.py
+      - docs/sparse-data.md
+      - examples/sparse_examples.py
+      - examples/render_readme_progression.py
+  verification:
+    automated:
+      - id: source-semantics
+        description: Preserve unknown periods and require explicit empty-volume
+          authorization.
+        command: python -m pytest -q tests/test_sparse.py -k "zero or omitted or grid or
+          duplicate or expansion"
+      - id: causal-boundaries
+        description: Retain real phase, event times, resets and as-of revision eligibility.
+        command: python -m pytest -q tests/test_sparse.py -k "phase or gap or reset or
+          event or revision or arrival"
+      - id: public-interfaces
+        description: Preserve useful rules, relationships, views, cases and API parity.
+        command: python -m pytest -q tests/test_sparse.py -k "rule or relationship or
+          daily or parity"
+      - id: progressive-examples
+        description: Reproduce the full numbered sequence and sparse worked examples.
+        command: python examples/run_demo.py
+    manual:
+      - Inspect progression figures 14–22 for true gaps, distinguished policy
+        zeros, original event time, unavailable checks and caller-owned business
+        severity.
+  review:
+    depth: partial
+    source: deterministic-assessment
+    gaps:
+      - Bounded calendar preparation preserves phase, original values and
+        coverage; only an explicit additive-volume empty policy authorizes zero
+        conversion. Review the cited evidence. If it is correct, save
+        agent.review evidence or make the criterion/reference wording more
+        explicit.
+      - Frozen eligible training and calibration support point references across
+        gaps, while pattern and change comparisons respect gaps and explicit
+        resets. Review the cited evidence. If it is correct, save agent.review
+        evidence or make the criterion/reference wording more explicit.
+      - Event-index profiles preserve real event timestamps and reject calendar
+        seasonality; unavailable relationship inputs and zero denominators stay
+        unavailable. Review the cited evidence. If it is correct, save
+        agent.review evidence or make the criterion/reference wording more
+        explicit.
+      - As-of replay respects supplied arrival/revision timestamps and sparse
+        multi-resolution views retain useful siblings without inventing complete
+        totals. Review the cited evidence. If it is correct, save agent.review
+        evidence or make the criterion/reference wording more explicit.
+    intent_summary: Implementation evidence synchronized from deterministic
+      assessment for data/quality/analyze-sparse-observations.
+    criteria:
+      - criterion: Bounded calendar preparation preserves phase, original values and
+          coverage; only an explicit additive-volume empty policy authorizes
+          zero conversion.
+        status: partial
+        evidence:
+          - src/anomalyzer/contracts.py:51
+          - src/anomalyzer/contracts.py:54
+          - src/anomalyzer/contracts.py:71
+          - src/anomalyzer/contracts.py:85
+          - src/anomalyzer/contracts.py:86
+          - src/anomalyzer/sparse.py:3
+          - src/anomalyzer/sparse.py:34
+          - src/anomalyzer/sparse.py:42
+          - src/anomalyzer/sparse.py:48
+          - src/anomalyzer/sparse.py:54
+          - src/anomalyzer/sparse_views.py:74
+          - src/anomalyzer/sparse_views.py:81
+          - src/anomalyzer/sparse_views.py:82
+          - src/anomalyzer/sparse_views.py:97
+          - src/anomalyzer/sparse_views.py:119
+          - src/anomalyzer/orchestrator.py:63
+          - src/anomalyzer/orchestrator.py:127
+          - src/anomalyzer/orchestrator.py:139
+          - src/anomalyzer/orchestrator.py:191
+          - src/anomalyzer/relationships.py:36
+          - src/anomalyzer/relationships.py:58
+          - src/anomalyzer/relationships.py:98
+          - src/anomalyzer/relationships.py:131
+          - src/anomalyzer/relationships.py:141
+          - src/anomalyzer/early.py:58
+          - src/anomalyzer/early.py:137
+          - tests/test_sparse.py:44
+          - tests/test_sparse.py:45
+          - tests/test_sparse.py:50
+          - tests/test_sparse.py:51
+          - tests/test_sparse.py:57
+          - docs/sparse-data.md:4
+          - docs/sparse-data.md:29
+          - docs/sparse-data.md:37
+          - docs/sparse-data.md:39
+          - docs/sparse-data.md:40
+          - examples/sparse_examples.py:22
+          - examples/sparse_examples.py:23
+          - examples/sparse_examples.py:25
+          - examples/sparse_examples.py:26
+          - examples/sparse_examples.py:28
+          - examples/render_readme_progression.py:135
+          - examples/render_readme_progression.py:139
+          - examples/render_readme_progression.py:140
+          - examples/render_readme_progression.py:166
+          - examples/render_readme_progression.py:206
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: Frozen eligible training and calibration support point references
+          across gaps, while pattern and change comparisons respect gaps and
+          explicit resets.
+        status: partial
+        evidence:
+          - src/anomalyzer/contracts.py:27
+          - src/anomalyzer/contracts.py:49
+          - src/anomalyzer/contracts.py:51
+          - src/anomalyzer/contracts.py:54
+          - src/anomalyzer/contracts.py:56
+          - src/anomalyzer/sparse.py:1
+          - src/anomalyzer/sparse.py:23
+          - src/anomalyzer/sparse.py:181
+          - src/anomalyzer/sparse.py:187
+          - src/anomalyzer/sparse.py:189
+          - src/anomalyzer/sparse_views.py:15
+          - src/anomalyzer/sparse_views.py:98
+          - src/anomalyzer/sparse_views.py:107
+          - src/anomalyzer/sparse_views.py:108
+          - src/anomalyzer/sparse_views.py:112
+          - src/anomalyzer/orchestrator.py:22
+          - src/anomalyzer/orchestrator.py:23
+          - src/anomalyzer/orchestrator.py:38
+          - src/anomalyzer/orchestrator.py:129
+          - src/anomalyzer/orchestrator.py:141
+          - src/anomalyzer/relationships.py:185
+          - src/anomalyzer/relationships.py:189
+          - src/anomalyzer/relationships.py:191
+          - src/anomalyzer/early.py:46
+          - src/anomalyzer/early.py:48
+          - src/anomalyzer/early.py:51
+          - src/anomalyzer/early.py:59
+          - src/anomalyzer/early.py:61
+          - tests/test_sparse.py:20
+          - tests/test_sparse.py:108
+          - tests/test_sparse.py:111
+          - tests/test_sparse.py:113
+          - tests/test_sparse.py:119
+          - docs/sparse-data.md:37
+          - docs/sparse-data.md:41
+          - docs/sparse-data.md:42
+          - docs/sparse-data.md:106
+          - docs/sparse-data.md:108
+          - examples/sparse_examples.py:16
+          - examples/sparse_examples.py:33
+          - examples/sparse_examples.py:50
+          - examples/sparse_examples.py:51
+          - examples/sparse_examples.py:91
+          - examples/render_readme_progression.py:241
+          - examples/render_readme_progression.py:312
+          - examples/render_readme_progression.py:317
+          - examples/render_readme_progression.py:346
+          - examples/render_readme_progression.py:349
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: Event-index profiles preserve real event timestamps and reject
+          calendar seasonality; unavailable relationship inputs and zero
+          denominators stay unavailable.
+        status: partial
+        evidence:
+          - src/anomalyzer/contracts.py:54
+          - src/anomalyzer/contracts.py:81
+          - src/anomalyzer/contracts.py:242
+          - src/anomalyzer/contracts.py:245
+          - src/anomalyzer/contracts.py:258
+          - src/anomalyzer/sparse.py:4
+          - src/anomalyzer/sparse.py:49
+          - src/anomalyzer/sparse.py:57
+          - src/anomalyzer/sparse.py:63
+          - src/anomalyzer/sparse.py:64
+          - src/anomalyzer/sparse_views.py:75
+          - src/anomalyzer/sparse_views.py:94
+          - src/anomalyzer/sparse_views.py:127
+          - src/anomalyzer/sparse_views.py:155
+          - src/anomalyzer/sparse_views.py:157
+          - src/anomalyzer/orchestrator.py:37
+          - src/anomalyzer/orchestrator.py:38
+          - src/anomalyzer/orchestrator.py:166
+          - src/anomalyzer/orchestrator.py:178
+          - src/anomalyzer/orchestrator.py:180
+          - src/anomalyzer/relationships.py:40
+          - src/anomalyzer/relationships.py:57
+          - src/anomalyzer/relationships.py:58
+          - src/anomalyzer/relationships.py:96
+          - src/anomalyzer/relationships.py:97
+          - src/anomalyzer/early.py:140
+          - src/anomalyzer/early.py:147
+          - tests/test_sparse.py:59
+          - tests/test_sparse.py:67
+          - tests/test_sparse.py:75
+          - tests/test_sparse.py:99
+          - tests/test_sparse.py:128
+          - docs/sparse-data.md:26
+          - docs/sparse-data.md:96
+          - docs/sparse-data.md:101
+          - docs/sparse-data.md:132
+          - docs/sparse-data.md:142
+          - examples/sparse_examples.py:26
+          - examples/sparse_examples.py:36
+          - examples/sparse_examples.py:38
+          - examples/sparse_examples.py:57
+          - examples/sparse_examples.py:81
+          - examples/render_readme_progression.py:654
+          - examples/render_readme_progression.py:697
+          - examples/render_readme_progression.py:884
+          - examples/render_readme_progression.py:896
+          - examples/render_readme_progression.py:917
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: As-of replay respects supplied arrival/revision timestamps and sparse
+          multi-resolution views retain useful siblings without inventing
+          complete totals.
+        status: partial
+        evidence:
+          - src/anomalyzer/contracts.py:13
+          - src/anomalyzer/contracts.py:54
+          - src/anomalyzer/contracts.py:297
+          - src/anomalyzer/contracts.py:298
+          - src/anomalyzer/contracts.py:437
+          - src/anomalyzer/sparse.py:4
+          - src/anomalyzer/sparse.py:34
+          - src/anomalyzer/sparse.py:64
+          - src/anomalyzer/sparse.py:66
+          - src/anomalyzer/sparse.py:83
+          - src/anomalyzer/sparse_views.py:1
+          - src/anomalyzer/sparse_views.py:28
+          - src/anomalyzer/sparse_views.py:29
+          - src/anomalyzer/sparse_views.py:82
+          - src/anomalyzer/sparse_views.py:94
+          - src/anomalyzer/orchestrator.py:106
+          - src/anomalyzer/orchestrator.py:107
+          - src/anomalyzer/orchestrator.py:133
+          - src/anomalyzer/orchestrator.py:166
+          - src/anomalyzer/orchestrator.py:180
+          - src/anomalyzer/relationships.py:160
+          - src/anomalyzer/relationships.py:162
+          - src/anomalyzer/relationships.py:166
+          - tests/test_sparse.py:59
+          - tests/test_sparse.py:75
+          - tests/test_sparse.py:99
+          - tests/test_sparse.py:123
+          - tests/test_sparse.py:180
+          - docs/sparse-data.md:26
+          - docs/sparse-data.md:41
+          - docs/sparse-data.md:69
+          - docs/sparse-data.md:78
+          - docs/sparse-data.md:96
+          - examples/sparse_examples.py:59
+          - examples/sparse_examples.py:95
+          - examples/render_readme_progression.py:686
+          - examples/render_readme_progression.py:699
+          - examples/render_readme_progression.py:710
+          - examples/render_readme_progression.py:767
+          - examples/render_readme_progression.py:884
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+    done: false
diff --git a/.capabilities/data/quality/exclude-incomplete-periods.capability.yaml b/.capabilities/data/quality/exclude-incomplete-periods.capability.yaml
index 5acd8a1..19ef979 100644
--- a/.capabilities/data/quality/exclude-incomplete-periods.capability.yaml
+++ b/.capabilities/data/quality/exclude-incomplete-periods.capability.yaml
@@ -1,11 +1,11 @@
 title: Avoid mistaking unfinished periods for drops
 status: implemented
-summary: Explicitly incomplete trailing observations are reported as data
-  quality evidence and excluded from completed-series analysis.
+summary: Explicitly incomplete observations are reported as data quality
+  evidence and excluded from completed-series analysis.
 intent: Prevent an unfinished period from appearing to be an unexplained drop.
 acceptance:
-  - Incomplete observations form a trailing suffix after chronological
-    preparation.
+  - Schema-1.0 incomplete observations form a trailing suffix; schema-1.1 source
+    policies preserve internal and trailing incomplete positions as unavailable.
   - The observed values and timestamps of unfinished periods remain available in
     data quality output.
   - Incomplete source samples are excluded from expectations, calibration, point
@@ -36,6 +36,9 @@ agent:
     - data/quality/prepare-timestamped-series
   implementation:
     references:
+      - src/anomalyzer/sparse.py
+      - src/anomalyzer/sparse_views.py
+      - tests/test_sparse.py
       - src/anomalyzer/validation.py
       - src/anomalyzer/alignment.py
       - src/anomalyzer/recipes.py
diff --git a/.capabilities/data/quality/prepare-timestamped-series.capability.yaml b/.capabilities/data/quality/prepare-timestamped-series.capability.yaml
index ea9da81..616b4e1 100644
--- a/.capabilities/data/quality/prepare-timestamped-series.capability.yaml
+++ b/.capabilities/data/quality/prepare-timestamped-series.capability.yaml
@@ -4,8 +4,9 @@ summary: Timestamped observations are normalized to UTC, ordered, and checked
   against their declared cadence before analysis.
 intent: Compare like periods and make preparation decisions visible to the caller.
 acceptance:
-  - Timestamped datasets provide both timestamps and frequency; positional
-    datasets provide neither.
+  - Calendar datasets provide timestamps and frequency; positional datasets
+    provide neither. Explicit schema-1.1 event-index sources retain timestamps
+    without a calendar frequency.
   - Local timestamps require a timezone, and ambiguous or nonexistent
     daylight-saving times require an explicit offset.
   - Duplicate timestamps are rejected unless an aggregation policy is supplied;
@@ -13,7 +14,9 @@ acceptance:
   - An as_of cutoff excludes later observations, and cadence is checked as a
     fixed elapsed duration.
 guidance:
-  - No interpolation, imputation, resampling, or unit conversion occurs.
+  - No interpolation or implicit filling, resampling, or unit conversion occurs.
+    Schema-1.1 source policies may explicitly authorize confirmed empty additive
+    volume as zero.
   - Duplicate null values propagate through aggregation.
 planning:
   story_map:
@@ -34,6 +37,8 @@ agent:
   # -----------------------------
   implementation:
     references:
+      - src/anomalyzer/sparse.py
+      - tests/test_sparse.py
       - src/anomalyzer/validation.py
       - src/anomalyzer/contracts.py
       - tests/test_analysis.py
diff --git a/.capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml b/.capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml
index dc06322..fc5b149 100644
--- a/.capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml
+++ b/.capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml
@@ -14,6 +14,10 @@ acceptance:
   - Invalid or conflicting revisions leave the previous state available;
     unchanged inputs preserve numerical evidence and unrelated source results.
 guidance:
+  - Schema-1.1 snapshot revision selection and as-of availability are
+    implemented in data/quality/analyze-sparse-observations; durable prior/new
+    evidence diffing and automatic reconciliation in this capability remain
+    planned.
   - A corrected event-time result does not replace the record of what was known
     earlier; callers control retention and any notification about a revision.
   - Proposed automated checks cover late completion, corrected calibration data,
diff --git a/.capabilities/dependency-graph.svg b/.capabilities/dependency-graph.svg
index 4c6b1df..1a36f80 100644
--- a/.capabilities/dependency-graph.svg
+++ b/.capabilities/dependency-graph.svg
@@ -105,7 +105,7 @@
   </g>
   <text class="drag-note" x="52" y="950">Open this SVG directly in a browser for animation and dragging.</text>
   <script><![CDATA[
-const graph = {"nodes":[{"id":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","label":"Understand how much histor...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","acceptance":["Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","Provisional residual scores can be reported but do not trigger point-anomaly episodes.","Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger."],"guidance":["Evidence strength is not a probability of an incident.","Calibration and expectations are separate concerns; each declared regime calibrates again."],"path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/detection/identify-residual-patterns","analysis/detection/report-point-anomaly-episodes","analysis/evidence/assess-reference-readiness","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","relationships/comparisons/compare-declared-metrics"],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"short-history","description":"Keep reference-only evidence for a short series.","command":"python -m pytest -q tests/test_analysis.py::test_short_series_returns_reference_evidence"},{"id":"provisional-evidence","description":"Report provisional scores without point triggers.","command":"python -m pytest -q tests/test_analysis.py::test_early_evidence_maturity_and_no_provisional_trigger"},{"id":"independent-calibration","description":"Check a known residual center, sample scale, and standardized score.","command":"python -m pytest -q tests/test_analysis.py::test_calibration_reference"},{"id":"constant-reference","description":"Handle zero residual variance without spurious episodes.","command":"python -m pytest -q tests/test_analysis.py::test_constant_and_schema"}],"manualChecks":["Inspect a sample's signal_maturity, calibration_samples, standardized_residual, and triggers together before describing a statistical departure."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","criteria":[{"criterion":"Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","status":"covered","evidence":["src/anomalyzer/recipes.py:205","src/anomalyzer/recipes.py:445","tests/test_analysis.py:141"],"notes":"The short-series test checks three reference-only samples, absent standardized scores, and no triggers."},{"criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"covered","evidence":["src/anomalyzer/recipes.py:210","tests/test_analysis.py:123"],"notes":"A provisional score of 2 exceeds the configured point threshold of 1 but produces no trigger or episode."},{"criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"covered","evidence":["src/anomalyzer/recipes.py:231","src/anomalyzer/recipes.py:307","tests/test_analysis.py:111"],"notes":"Calibration uses retained residual mean and sample standard deviation once per regime; the independent numerical example produces the expected standardized score."},{"criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"covered","evidence":["src/anomalyzer/contracts.py:104","src/anomalyzer/recipes.py:350","src/anomalyzer/recipes.py:388","tests/test_analysis.py:111"],"notes":"Sample fields retain the comparison and support; the point threshold is inspectable in the enclosing method parameters."}],"ignore_findings":[],"done":true,"input_fingerprint":"ccdc1b2e92f3e647d5a9e6127cc3c4b10bf0da8f1f78af3efb805daa9b1e4305","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Establish evidence support"},"impact":6,"gaps":0,"r":68,"x":170,"y":170},{"id":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","label":"Bring my metrics into an a...","area":"data","scopes":["data","data/inputs"],"scope":"data/inputs","scopeLabel":"data/inputs","status":"implemented","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","acceptance":["Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","Published schema 1.0 and 1.1 request and result schemas match the public Python contracts."],"guidance":["Null observations represent missing values, never zero. Limits apply to the request as a whole.","Describe current supported behavior; historical plans do not add accepted inputs."],"path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["analysis/evidence/calibrate-departures","analysis/expectations/model-seasonal-level-and-growth","execution/resilience/retain-evidence-under-failure-and-limits","interfaces/python/analyze-through-public-api","investigation/assessments/describe-early-changes"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_agentic.py","schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json"],"automatedChecks":[{"id":"input-forms","description":"Analyze a bare list and validate a canonical request and result.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings tests/test_analysis.py::test_constant_and_schema"},{"id":"explicit-scope","description":"Reject schema 1.1 datasets without entity scope.","command":"python -m pytest -q tests/test_agentic.py::test_schema_11_requires_explicit_entity_scope"},{"id":"invalid-inputs","description":"Reject unsupported settings and nonnumeric or nonfinite observations.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_config tests/test_analysis.py::test_invalid_values"},{"id":"published-schemas","description":"Compare all four checked-in schemas with their public contracts.","command":"python -m pytest -q tests/test_analysis.py::test_checked_in_schemas tests/test_agentic.py::test_v11_checked_in_schemas_exist"}],"manualChecks":["Review the input examples in README.md and confirm they describe the two supported schema versions and the four-dataset/four-relationship limit."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","criteria":[{"criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"covered","evidence":["src/anomalyzer/recipes.py:21","src/anomalyzer/contracts.py:96","tests/test_analysis.py:234"],"notes":"Bare-list settings are validated; schema 1.0 limits datasets to exactly one."},{"criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"covered","evidence":["src/anomalyzer/contracts.py:240","src/anomalyzer/contracts.py:335","tests/test_agentic.py:192"],"notes":"Entity scope is required, relationship kinds are discriminated, and references must name distinct known datasets."},{"criterion":"Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:7","src/anomalyzer/contracts.py:71","tests/test_analysis.py:526","tests/test_analysis.py:531"],"notes":"Assertions reject each listed category; missing null values remain a separate accepted representation."},{"criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"covered","evidence":["schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json","tests/test_analysis.py:547","tests/test_agentic.py:317"],"notes":"Tests and independent inspection compare each complete published schema with its contract-generated schema. Python cross-field validators supplement JSON Schema constraints."}],"ignore_findings":[],"done":true,"input_fingerprint":"3546cd8f436812ff11c9352a855926677b55cc6287e40f6a5041bd613173e5b4","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Supply a numeric series"},"impact":5,"gaps":0,"r":63,"x":353.33333333333337,"y":170},{"id":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","label":"Analyze local files from t...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","acceptance":["The methods command lists supported analysis recipes and residual detectors.","Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status."],"guidance":["CSV callers must select one entity and unit before analysis.","Exit 0 includes completed analysis with anomalies, insufficient history, or inapplicability; exit 2 means invalid input, and exit 1 means execution failure or partial execution."],"path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/python/install-versioned-library","interfaces/skill/guide-anomaly-investigation","interfaces/skill/visualize-analysis-on-request"],"implementationReferences":["src/anomalyzer/cli.py","src/anomalyzer/io.py","src/anomalyzer/__main__.py","tests/test_cli.py"],"automatedChecks":[{"id":"file-stdin","description":"Analyze equivalent file and stdin requests.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence"},{"id":"csv-precedence","description":"Validate selected CSV fields and apply flag-over-file-over-embedded precedence.","command":"python -m pytest -q tests/test_cli.py::test_csv_and_precedence"},{"id":"value-list","description":"Analyze a bare JSON array and expose inferred settings.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"},{"id":"output-and-exits","description":"Protect existing output and distinguish invalid options from failed execution.","command":"python -m pytest -q tests/test_cli.py::test_output_protection tests/test_cli.py::test_methods_and_invalid_options tests/test_cli.py::test_size_limit_and_execution_exit"}],"manualChecks":["Run an example in readable and JSON modes and confirm stdout contains results while diagnostics go to stderr."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","criteria":[{"criterion":"The methods command lists supported analysis recipes and residual detectors.","status":"covered","evidence":["src/anomalyzer/cli.py:426","src/anomalyzer/registry.py:15","tests/test_cli.py:27"],"notes":"Discovery tests check all three recipes and ten residual detectors, with structured stdout and no successful-run diagnostics on stderr."},{"criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:239","tests/test_cli.py:13","tests/test_mcp.py:163"],"notes":"The passing file/stdin regression compares request fingerprints and short-history semantics; cross-interface tests compare complete stable numerical output."},{"criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"covered","evidence":["src/anomalyzer/io.py:33","tests/test_cli.py:51"],"notes":"The passing CSV regression covers timestamped missing cells, positional data, and varying unmapped entity rejection."},{"criterion":"Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","status":"covered","evidence":["src/anomalyzer/cli.py:253","tests/test_cli.py:51","tests/test_cli.py:71"],"notes":"The passing file/flag/embedded precedence regression and bare-values regression establish the merge order and validated configuration."},{"criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"covered","evidence":["src/anomalyzer/io.py:69","src/anomalyzer/cli.py:424","tests/test_cli.py:41","tests/test_cli.py:108"],"notes":"Output protection and exit regressions passed, including explicit overwrite; successful manual examples have empty stderr and incomplete-usage exits 1."}],"ignore_findings":[],"done":true,"input_fingerprint":"86e701ccc0a1aefd407f11b4547784fef497fa1b49ddf6bb69db0d08f757dfdc","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Run from files or stdin"},"impact":4,"gaps":0,"r":58,"x":903.3333333333334,"y":170},{"id":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","label":"Account for ordinary seaso...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","acceptance":["A supplied season length selects earlier matching seasonal observations for the reference.","Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected."],"guidance":["Expectations support scoring; they are not a standalone forecasting API."],"path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/expectations/protect-reference-from-extremes","analysis/expectations/reset-declared-regimes","analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py"],"automatedChecks":[{"id":"seasonal-reference","description":"Check independent lagged-reference values.","command":"python -m pytest -q tests/test_analysis.py::test_forecast_independent_reference"},{"id":"growth-and-departures","description":"Keep clean linear and compound growth ordinary while identifying injected departures.","command":"python -m pytest -q tests/test_trend.py::test_trend_and_anomalies"},{"id":"frozen-fit","description":"Keep the training fit stable when later observations change.","command":"python -m pytest -q tests/test_trend.py::test_frozen_fit_and_legacy_mode"},{"id":"explicit-trends","description":"Accept supported trend choices and handle short, constant, and signed series.","command":"python -m pytest -q tests/test_trend.py::test_explicit_modes_and_validation tests/test_trend.py::test_short_history_constant_and_signed_linear"}],"manualChecks":["Compare examples/growth_up.json and examples/compound.json with their selected trend and evidence expectations."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","criteria":[{"criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"covered","evidence":["src/anomalyzer/baseline.py:8","src/anomalyzer/baseline.py:31","tests/test_analysis.py:21"],"notes":"Independent lag values and robust matching-phase assertions establish the seasonal reference calculation."},{"criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"covered","evidence":["src/anomalyzer/trend.py:120","src/anomalyzer/trend.py:144","tests/test_trend.py:11"],"notes":"Positive and negative linear/exponential cases select the expected model, match post-training expectations, and produce no ordinary-growth episodes."},{"criterion":"The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","status":"covered","evidence":["src/anomalyzer/recipes.py:162","tests/test_trend.py:36"],"notes":"Trend fitting occurs at the training boundary; the test asserts identical trend diagnostics and earlier evidence after later observations change."},{"criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:20","tests/test_trend.py:49","tests/test_trend.py:57"],"notes":"Tests exercise disabled, linear, exponential, and invalid quadratic modes, including disclosed short-history fallback."}],"ignore_findings":[],"done":true,"input_fingerprint":"8e9a0361b02b7d3f651366e5fa0675cccb3949e44de448d2ae0d43e6710bd408","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Account for growth and seasonality"},"impact":3,"gaps":0,"r":53,"x":170,"y":170},{"id":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","label":"Embed local analysis in my...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","acceptance":["analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","analyze_series preserves supplied settings and event-time context for a single dataset.","analyze_relationships preserves the declared multi-dataset relationship semantics.","Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime."],"guidance":["All interfaces use local analysis; there is no hosted service or model API requirement."],"path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence","interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/__init__.py","src/anomalyzer/agent.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"request-parity","description":"Match canonical request results across Python, CLI, and full-result MCP compatibility calls.","command":"python -m pytest -q tests/test_mcp.py::test_canonical_request_matches_python_cli_and_mcp tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"series-context","description":"Preserve a single-series event-time cutoff and context across interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp tests/test_mcp.py::test_series_wrapper_preserves_omitted_and_explicit_season_settings"},{"id":"relationship-parity","description":"Match declared relationship evidence across the public interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_relationship_semantics_match_python_cli_and_mcp"}],"manualChecks":["Review the Python API examples in README.md and confirm callers can inspect the returned validated models and JSON."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","criteria":[{"criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"covered","evidence":["src/anomalyzer/__init__.py:3","src/anomalyzer/recipes.py:21","tests/test_analysis.py:234"],"notes":"Bare values and separate settings validate before dispatch; canonical schema 1.0 and 1.1 requests use their respective contracts."},{"criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"covered","evidence":["src/anomalyzer/agent.py:11","tests/test_mcp.py:191","tests/test_mcp.py:291"],"notes":"Field-set handling preserves omitted versus explicit settings and context in the single-series wrapper, verified across dictionary and Settings inputs."},{"criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"covered","evidence":["src/anomalyzer/agent.py:27","tests/test_mcp.py:217"],"notes":"Declared relationships validate and use the shared engine without changing units, lineage, or primitive semantics."},{"criterion":"Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime.","status":"covered","evidence":["tests/test_mcp.py:163","tests/test_mcp.py:191","tests/test_mcp.py:217","src/anomalyzer/__main__.py:1"],"notes":"Passing real CLI subprocess, Python, and full-result MCP parity tests compare stable numerical results apart from identity and runtime."}],"ignore_findings":[],"done":true,"input_fingerprint":"214bd10bb3aa5fad0f74938fe2ef09853a593cdb5b93cba49ec851a628d8ab65","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","name":"Python integration","label":"Python integration / Embed analysis in my application / Call the public API"},"impact":3,"gaps":0,"r":53,"x":903.3333333333334,"y":170},{"id":"investigation/cases/group-related-evidence","title":"Investigate related findings together","label":"Investigate related findin...","area":"investigation","scopes":["investigation","investigation/cases"],"scope":"investigation/cases","scopeLabel":"investigation/cases","status":"implemented","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","acceptance":["Evidence for different declared entities remains in separate cases even at the same event time.","Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","A derived relationship adds context without counting as an additional independent source for severity.","Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection."],"guidance":["Warning and critical are deterministic numerical grouping rules, not validated incident impact.","Diagnostic trend, oscillation, mixture, and variation findings remain distinct from adaptive location-shift cases."],"path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","dependencies":["investigation/assessments/describe-early-changes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/mcp/investigate-findings-and-evidence","investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/cases.py","src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py"],"automatedChecks":[{"id":"entity-scope","description":"Separate cases for different entities at the same event time.","command":"python -m pytest -q tests/test_agentic.py::test_cases_are_scoped_by_entity_at_the_same_event_time"},{"id":"source-severity","description":"Escalate numerical severity when two source datasets meet criteria.","command":"python -m pytest -q tests/test_agentic.py::test_two_source_datasets_escalate_case_severity"},{"id":"correlated-context","description":"Keep relationship evidence from double-counting source severity.","command":"python -m pytest -q tests/test_agentic.py::test_relationship_evidence_explains_but_does_not_double_count_severity"},{"id":"first-detection","description":"Group an adaptive location shift before point triggers without using a later pattern peak.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_location_shift_creates_one_early_case_without_point_trigger tests/test_agentic.py::test_location_case_severity_does_not_use_future_pattern_peak"}],"manualChecks":["Inspect a case's contributing targets, explanation, evidence references, and investigation questions; distinguish it from a point-anomaly episode."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","criteria":[{"criterion":"Evidence for different declared entities remains in separate cases even at the same event time.","status":"covered","evidence":["src/anomalyzer/cases.py:42","tests/test_agentic.py:69"],"notes":"Entity participates in grouping and stable identity; the test requires separate cases and distinct IDs at the same time."},{"criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"covered","evidence":["src/anomalyzer/cases.py:72","tests/test_agentic.py:85"],"notes":"Two qualifying source datasets produce critical numerical severity and named source explanations."},{"criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"covered","evidence":["src/anomalyzer/cases.py:72","src/anomalyzer/cases.py:116","tests/test_agentic.py:100"],"notes":"The independent-source count includes dataset criteria only. Derived evidence remains explanatory and can still meet the separately declared magnitude rule."},{"criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"covered","evidence":["src/anomalyzer/early.py:193","src/anomalyzer/early.py:237","src/anomalyzer/orchestrator.py:115","tests/test_agentic.py:122","tests/test_agentic.py:146"],"notes":"Tests require one moderate-shift case without point triggers and restrict assessment references/strength to samples at or before first detection."}],"ignore_findings":[],"done":true,"input_fingerprint":"6830af8a2f064123c12adb062f4793e446b97b3b805fd73a934ea84cd0d5d9cf","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Investigate related evidence"},"impact":3,"gaps":0,"r":53,"x":1086.6666666666667,"y":170},{"id":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","label":"Analyze observations in th...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","acceptance":["Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration."],"guidance":["No interpolation, imputation, resampling, or unit conversion occurs.","Duplicate null values propagate through aggregation."],"path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["data/quality/exclude-incomplete-periods","relationships/quality/require-compatible-sources"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/contracts.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"coordinate-contract","description":"Reject partial timestamp metadata and positional as_of cutoffs.","command":"python -m pytest -q tests/test_analysis.py::test_timestamp_metadata_is_all_or_nothing"},{"id":"local-time","description":"Normalize valid local times and reject ambiguous or nonexistent times.","command":"python -m pytest -q tests/test_analysis.py::test_local_time"},{"id":"duplicates-cutoff","description":"Require an explicit duplicate policy and retain counts after the cutoff.","command":"python -m pytest -q tests/test_analysis.py::test_duplicate_sort_and_cutoff"},{"id":"cadence-context","description":"Check fixed weekly cadence and preserve cutoff semantics across interfaces.","command":"python -m pytest -q tests/test_analysis.py::test_weekly_cadence tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp"}],"manualChecks":["Inspect a timestamped result's data_quality transformations, UTC window, and resolved context to confirm preparation is explainable."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","criteria":[{"criterion":"Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","status":"covered","evidence":["src/anomalyzer/contracts.py:77","tests/test_analysis.py:418"],"notes":"Partial coordinate metadata and positional as_of are rejected."},{"criterion":"Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","status":"covered","evidence":["src/anomalyzer/validation.py:8","tests/test_analysis.py:500"],"notes":"UTC round trips detect ambiguous/nonexistent local times; valid local time normalizes to the expected UTC hour."},{"criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"covered","evidence":["src/anomalyzer/validation.py:87","src/anomalyzer/validation.py:103","tests/test_analysis.py:478"],"notes":"Explicit mean/sum handling retains duplicate counts and transformation descriptions. Nulls propagate rather than becoming zero."},{"criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"covered","evidence":["src/anomalyzer/validation.py:105","src/anomalyzer/validation.py:111","tests/test_analysis.py:491","tests/test_mcp.py:191"],"notes":"Filtering is inclusive through as_of; cadence compares UTC elapsed timedeltas, including fixed seven-day weeks."}],"ignore_findings":[],"done":true,"input_fingerprint":"791748764cdc0dfe81d987a908179930d858670606e51b64407172295776b261","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Prepare the comparison"},"impact":2,"gaps":0,"r":48,"x":353.33333333333337,"y":170},{"id":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","label":"Let an agent inspect the e...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","acceptance":["The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","get_evidence exposes the selected finding's numerical support and relationship source lineage.","Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated."],"guidance":["Run results are retained only in the bounded server-session cache; missing or evicted run IDs require reanalysis.","Cases, episodes, and residual patterns may overlap."],"path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":["interfaces/mcp/connect-read-only-local-server","interfaces/skill/guide-anomaly-investigation"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py"],"automatedChecks":[{"id":"case-drilldown","description":"Analyze, page findings, and inspect exact conversion case values and lineage.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case tests/test_mcp.py::test_schema_11_episode_is_paged_and_has_relationship_evidence"},{"id":"unavailable-finding","description":"Expose a quality finding when latest relationship evidence is unavailable.","command":"python -m pytest -q tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"},{"id":"early-candidate","description":"Inspect descriptive candidate evidence without inventing a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case tests/test_mcp.py::test_inapplicable_method_keeps_descriptive_candidate_visible"},{"id":"sample-pagination","description":"Page supporting samples for a residual pattern.","command":"python -m pytest -q tests/test_mcp.py::test_get_evidence_pages_pattern_samples"}],"manualChecks":["Follow analyze, list_findings, and get_evidence for the conversion example and check that the headline is supported by the retrieved values."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","criteria":[{"criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"covered","evidence":["src/anomalyzer/mcp.py:170","src/anomalyzer/mcp_views.py:247","tests/test_mcp.py:54"],"notes":"The conversion manual call returns a relationship headline, preview, counts, readiness, and session-scoped run ID."},{"criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:132","src/anomalyzer/mcp_views.py:260","tests/test_mcp.py:313","tests/test_mcp.py:108"],"notes":"Cases, early candidates, episodes, patterns, and quality findings feed a bounded page; relationship episode regression verifies schema 1.1 pagination."},{"criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:273","src/anomalyzer/mcp_views.py:294","tests/test_mcp.py:54","tests/test_mcp.py:313"],"notes":"Manual drilldown returns observed 8/130 versus expected 0.1 and source indexes orders=14 and qualified_visits=14."},{"criterion":"Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:118","src/anomalyzer/mcp_views.py:169","src/anomalyzer/mcp_views.py:337","tests/test_mcp.py:87","tests/test_mcp.py:137","tests/test_mcp.py:341"],"notes":"Passing tests retain short-history and inapplicable-method candidates plus unavailable latest evidence, while pattern samples are paginated."}],"ignore_findings":[],"done":true,"input_fingerprint":"a09b0003780bcd2f9ccf8e3d8ff58b3147c118760b984757d7027b5810fd9bef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Drill into findings and evidence"},"impact":2,"gaps":0,"r":48,"x":927.3333333333334,"y":288},{"id":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","label":"Spot early changes with a ...","area":"investigation","scopes":["investigation","investigation/assessments"],"scope":"investigation/assessments","scopeLabel":"investigation/assessments","status":"implemented","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","acceptance":["The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","Missing values break the comparison history, and zero previous values do not create relative-change evidence.","Early candidates remain non-triggering and ineligible for action even under a permissive policy."],"guidance":["An early candidate does not establish an anomaly, stable baseline, cause, or incident."],"path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence","investigation/rules/check-declared-numerical-rules"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_mcp.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"prefix-invariant-facts","description":"Describe short history causally without creating cases.","command":"python -m pytest -q tests/test_agentic.py::test_short_history_assessments_are_useful_and_prefix_invariant"},{"id":"comparison-boundaries","description":"Reset comparisons at missing observations and avoid dividing by zero.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"candidate-eligibility","description":"Keep candidates ineligible for action and notifications under permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"},{"id":"agent-visibility","description":"Expose a short-history candidate and its claim limits without a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case"}],"manualChecks":["Inspect a candidate's comparison population, maturity, established claim, and not_established text before describing it."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","criteria":[{"criterion":"The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","status":"covered","evidence":["src/anomalyzer/early.py:73","src/anomalyzer/early.py:96","tests/test_agentic.py:35"],"notes":"Assertions establish observation/change/change/departure_candidate for [10,12,11,20]."},{"criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"covered","evidence":["src/anomalyzer/early.py:50","src/anomalyzer/early.py:141","tests/test_agentic.py:35"],"notes":"Each assessment precedes the prior-history update; tests compare complete serialized assessments for every prefix."},{"criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"covered","evidence":["src/anomalyzer/early.py:53","src/anomalyzer/early.py:58","tests/test_capability_behaviors.py:87"],"notes":"The zero comparison has no relative change; after a missing value, comparison samples reset to zero."},{"criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"covered","evidence":["src/anomalyzer/early.py:105","src/anomalyzer/cases.py:35","tests/test_capability_behaviors.py:134","tests/test_mcp.py:341"],"notes":"Candidates have criterion_met=false and remain action/notification-ineligible after policy replay, without creating cases."}],"ignore_findings":[],"done":true,"input_fingerprint":"debdeeeaea559950f53461a4298cb9c2e3a1681f83a43e8a981ed9b51968a4d0","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect early evidence"},"impact":2,"gaps":0,"r":48,"x":1110.6666666666667,"y":288},{"id":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","label":"Compare the same entity an...","area":"relationships","scopes":["relationships","relationships/quality"],"scope":"relationships/quality","scopeLabel":"relationships/quality","status":"implemented","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","acceptance":["Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","Difference relationships require identical declared source units and matching output units.","An invalid relationship is reported as failed while valid sibling dataset results remain available."],"guidance":["A failed or unavailable relationship does not establish that its metric is normal.","No implicit timestamp matching, interpolation, unit conversion, or zero-filling occurs."],"path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"implementationReferences":["src/anomalyzer/alignment.py","src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"exact-alignment","description":"Reject mismatched timestamps, cadence, entity, duplicates, and positional relationship sources.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relationship_requires_exact_source_alignment"},{"id":"compatible-units","description":"Report a difference unit mismatch without suppressing sibling results.","command":"python -m pytest -q tests/test_agentic.py::test_alignment_unit_and_partial_failure_are_specific"},{"id":"unambiguous-declaration","description":"Reject fields that do not belong to the selected relationship kind.","command":"python -m pytest -q tests/test_agentic.py::test_contract_rejects_ambiguous_relationship_fields"},{"id":"healthy-siblings","description":"Retain a healthy dataset's evidence when another dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Read both target-level statuses and errors before interpreting a partial run; check source dataset errors when a relationship source failed."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","criteria":[{"criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"covered","evidence":["src/anomalyzer/alignment.py:42","tests/test_capability_behaviors.py:30"],"notes":"Parameterized assertions require specific errors for timestamp, cadence, and entity mismatches."},{"criterion":"Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","status":"covered","evidence":["src/anomalyzer/alignment.py:23","src/anomalyzer/alignment.py:44","tests/test_capability_behaviors.py:30"],"notes":"Positional sources fail exact alignment; even an explicit duplicate-mean setting is rejected for relationship source lineage."},{"criterion":"Difference relationships require identical declared source units and matching output units.","status":"covered","evidence":["src/anomalyzer/relationships.py:41","tests/test_agentic.py:242"],"notes":"Source and output unit checks are explicit. Tests cover mismatched source units and successful matching units."},{"criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:171","tests/test_capability_behaviors.py:30","tests/test_capability_behaviors.py:159"],"notes":"Relationship errors remain target-specific, healthy dataset results survive, and the overall mixed run is partial."}],"ignore_findings":[],"done":true,"input_fingerprint":"06321ab5806a8276cc3954df5d6f4dd7f059f67600038854569787ff51b420d6","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Align comparable sources"},"impact":2,"gaps":0,"r":48,"x":1294,"y":288},{"id":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","label":"Spot sustained shifts and ...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","acceptance":["Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","Trend, oscillation, and mixture findings retain their specific pattern meanings.","Moving-range findings identify variation changes separately from location changes."],"guidance":["Pattern findings may overlap and are not independent confirmations.","A residual location shift does not establish a change in the full distribution."],"path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence"],"implementationReferences":["src/anomalyzer/evidence.py","src/anomalyzer/recipes.py","tests/test_analysis.py"],"automatedChecks":[{"id":"location-shift","description":"Find moderate increases and decreases without point anomalies.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_patterns_describe_location_shift_without_point_anomaly"},{"id":"accumulating-change","description":"Detect a known CUSUM accumulation at the expected position.","command":"python -m pytest -q tests/test_analysis.py::test_cusum_detects_accumulated_moderate_location_shift"},{"id":"diagnostic-patterns","description":"Distinguish residual trend, oscillation, and mixture rules.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_diagnostics_have_specific_semantics"},{"id":"variation","description":"Describe a moving-range increase as variation rather than location.","command":"python -m pytest -q tests/test_analysis.py::test_moving_range_detects_short_term_variation_increase"}],"manualChecks":["Compare examples/location_shift_up.json with its point episodes and pattern findings, allowing for overlapping residual rules."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","criteria":[{"criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"covered","evidence":["src/anomalyzer/evidence.py:118","src/anomalyzer/evidence.py:189","tests/test_analysis.py:293"],"notes":"Tests assert no point triggers or episodes and exact upward/downward detection positions for Nelson rules and CUSUM."},{"criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"covered","evidence":["src/anomalyzer/evidence.py:211","src/anomalyzer/evidence.py:230","tests/test_analysis.py:350"],"notes":"The known accumulation detects at index 16 and reports peak CUSUM approximately 6, threshold 5, and increase direction."},{"criterion":"Trend, oscillation, and mixture findings retain their specific pattern meanings.","status":"covered","evidence":["src/anomalyzer/evidence.py:70","src/anomalyzer/evidence.py:122","tests/test_analysis.py:340"],"notes":"Separate predicates and descriptions preserve the three meanings and explicitly limit location/distribution claims."},{"criterion":"Moving-range findings identify variation changes separately from location changes.","status":"covered","evidence":["src/anomalyzer/evidence.py:234","tests/test_analysis.py:360"],"notes":"Adjacent standardized-residual differences produce variation_shift findings; the test checks magnitude, detection index, and interpretation."}],"ignore_findings":[],"done":true,"input_fingerprint":"63716eb67db6b73964b79897d8c81f2891df2586ff213c95ea81c5699b8abbea","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Find sustained residual changes"},"impact":1,"gaps":0,"r":43,"x":194,"y":288},{"id":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","label":"Avoid mistaking unfinished...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","acceptance":["Incomplete observations form a trailing suffix after chronological preparation.","The observed values and timestamps of unfinished periods remain available in data quality output.","Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","A relationship with an unavailable latest source reports that unavailability rather than normal behavior."],"guidance":["Completeness is declared by the caller; the engine does not infer collection completeness."],"path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/alignment.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"unfinished-values","description":"Retain an unfinished value without scoring it.","command":"python -m pytest -q tests/test_analysis.py::test_explicit_incomplete_period_is_reported_and_not_scored"},{"id":"trailing-only","description":"Reject incomplete observations followed by completed observations.","command":"python -m pytest -q tests/test_analysis.py::test_incomplete_periods_must_be_trailing"},{"id":"source-and-agent-views","description":"Exclude a partial source sample from both analysis views and expose unavailable relationship evidence.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"}],"manualChecks":["Inspect examples/agentic/incomplete-usage.json and confirm the latest headline explains missing completeness and does not imply normality."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","criteria":[{"criterion":"Incomplete observations form a trailing suffix after chronological preparation.","status":"covered","evidence":["src/anomalyzer/validation.py:44","tests/test_analysis.py:162"],"notes":"Suffix validation runs after timestamp preparation and rejects a completed observation following an incomplete one."},{"criterion":"The observed values and timestamps of unfinished periods remain available in data quality output.","status":"covered","evidence":["src/anomalyzer/validation.py:50","src/anomalyzer/validation.py:114","tests/test_analysis.py:150"],"notes":"Quality records retain original source index, normalized timestamp, observed value, and incomplete status."},{"criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"covered","evidence":["src/anomalyzer/validation.py:53","src/anomalyzer/recipes.py:564","tests/test_analysis.py:150","tests/test_analysis.py:200"],"notes":"Only finalized prepared values enter either scoring path; assertions verify the last scored sample and both multi-resolution counts."},{"criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"covered","evidence":["src/anomalyzer/alignment.py:35","src/anomalyzer/orchestrator.py:150","tests/test_mcp.py:87"],"notes":"Incomplete aligned sources become unavailable values, and latest relationship state/status plus MCP headline explain the lack of evidence."}],"ignore_findings":[],"done":true,"input_fingerprint":"d82ee816fe89a25232e2589587d31bcadfa81be4188a0e1531f4fbfdf0ebcb5e","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Keep unfinished data visible"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","label":"Give an agent read-only ac...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","acceptance":["Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","Analysis tool calls return structured content together with compact text.","Invalid tool input and unknown session run IDs return structured errors rather than fabricated results."],"guidance":["Compatibility tool names remain callable but are not advertised.","Legacy clients receive isError tool results; modern clients receive JSON-RPC invalid-input errors.","The server provides analysis and evidence inspection; the caller owns monitoring and operational workflows."],"path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","dependencies":["interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":["interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py","pyproject.toml"],"automatedChecks":[{"id":"tool-discovery","description":"Advertise the three read-only tools and their schema contracts.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_lists_read_only_tools"},{"id":"connection-protocols","description":"Support legacy stdio initialization and modern stateless discovery.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_stdio_legacy_handshake_and_tool_listing tests/test_mcp.py::test_mcp_modern_discovery_and_stateless_tool_listing"},{"id":"structured-content","description":"Return a compact analysis response with structured content and run ID.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_tool_call_returns_structured_content"},{"id":"invalid-calls","description":"Reject malformed input and unknown session run IDs in both protocols.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_invalid_tool_input_is_a_tool_error tests/test_mcp.py::test_paged_tool_inputs_enforce_schema_in_both_protocols"}],"manualChecks":["Review discovery's interpretation guidance and confirm the client can start anomalyzer-mcp in the installed Python environment."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","criteria":[{"criterion":"Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","status":"covered","evidence":["src/anomalyzer/mcp.py:63","tests/test_mcp.py:27"],"notes":"Discovery exposes three read-only tools with the same strict argument contracts used at dispatch and advertised output schemas."},{"criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"covered","evidence":["src/anomalyzer/mcp.py:228","tests/test_mcp.py:248","tests/test_mcp.py:268"],"notes":"Legacy initialization and modern stateless discovery tests establish protocol metadata and evidence interpretation guidance."},{"criterion":"Analysis tool calls return structured content together with compact text.","status":"covered","evidence":["src/anomalyzer/mcp.py:162","tests/test_mcp.py:39"],"notes":"The manual conversion call returns a compact headline and structured run ID, counts, readiness, and preview."},{"criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"covered","evidence":["src/anomalyzer/mcp.py:156","src/anomalyzer/mcp.py:257","tests/test_mcp.py:235","tests/test_mcp.py:367"],"notes":"Tests reject invalid cursor/limit types, unknown fields, numeric run IDs, and missing session runs in both protocols without fabricated results."}],"ignore_findings":[],"done":true,"input_fingerprint":"7bd92d238f4f9545372c6c35e78aefea60c585bea0a4a8bb99096ab6cef27f96","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Connect a local server"},"impact":1,"gaps":0,"r":43,"x":927.3333333333334,"y":288},{"id":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","label":"Find values that break my ...","area":"investigation","scopes":["investigation","investigation/rules"],"scope":"investigation/rules","scopeLabel":"investigation/rules","status":"implemented","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","acceptance":["Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","Absolute and relative changes must exceed their declared maximum to violate a rule.","Relative-change checks require a nonzero previous value and do not bridge missing observations.","Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them."],"guidance":["An exact rule crossing does not establish cause, an incident, or business impact."],"path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","dependencies":["investigation/assessments/describe-early-changes"],"suggestedDependencies":[],"dependents":["investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"rule-boundaries","description":"Check inclusive range boundaries and strict absolute/relative change thresholds.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_explicit_rules_respect_boundaries"},{"id":"relative-availability","description":"Avoid relative violations across zero or missing values.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"explicit-not-actionable","description":"Create evidence and a case for an exact rule crossing without enabling action by default.","command":"python -m pytest -q tests/test_agentic.py::test_explicit_rule_is_exact_but_not_actionable_without_policy"},{"id":"readable-rule","description":"Identify the exact unit-cost range rule in CLI output.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the unit-cost example's crossed rule and confirm its text distinguishes the declared bound from statistical detection."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","criteria":[{"criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"covered","evidence":["src/anomalyzer/early.py:22","src/anomalyzer/contracts.py:209","tests/test_capability_behaviors.py:74"],"notes":"Boundary values 10 and 20 are accepted; 9 and 21 violate the declared range."},{"criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"covered","evidence":["src/anomalyzer/early.py:25","tests/test_capability_behaviors.py:74"],"notes":"Parameterized tests distinguish equality to the threshold from strict exceedance for both change rules."},{"criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"covered","evidence":["src/anomalyzer/early.py:29","src/anomalyzer/early.py:53","tests/test_capability_behaviors.py:87"],"notes":"Zero/missing predecessors suppress relative comparisons, and missing values reset comparison history."},{"criterion":"Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them.","status":"covered","evidence":["src/anomalyzer/early.py:34","src/anomalyzer/early.py:121","tests/test_agentic.py:49","tests/test_cli.py:134"],"notes":"Violation records retain rule ID/kind/bounds; default eligibility is false and replay requires explicit permission."}],"ignore_findings":[],"done":true,"input_fingerprint":"60d24e84faee9c1e2ae8f9337d17334b37c89305093404da2e84c51526111d0b","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Apply my stated limits"},"impact":1,"gaps":0,"r":43,"x":1110.6666666666667,"y":288},{"id":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","label":"Find unusual spikes and dr...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","acceptance":["A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","Separated point departures remain separate episodes, and positional episode intervals use integer coordinates."],"guidance":["Detection is a numerical departure from the selected expectation, not a claim about cause or business impact."],"path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/evidence.py","tests/test_analysis.py"],"automatedChecks":[{"id":"point-evidence","description":"Detect an injected spike and retain deterministic earlier numerical evidence.","command":"python -m pytest -q tests/test_analysis.py::test_spike_and_no_lookahead"},{"id":"adjacent-run","description":"Describe adjacent point triggers and their contributing positions.","command":"python -m pytest -q tests/test_analysis.py::test_anomaly_stream_detects_consecutive_run"},{"id":"isolated-events","description":"Keep separated anomalies from becoming a consecutive-run pattern.","command":"python -m pytest -q tests/test_analysis.py::test_isolated_anomalies_do_not_form_second_order_pattern"},{"id":"positional-episodes","description":"Use integer intervals for position-only episode evidence.","command":"python -m pytest -q tests/test_analysis.py::test_positional_episode_uses_integer_interval"}],"manualChecks":["Inspect observations and triggering_samples for examples/spike.json and distinguish episode grouping from pattern findings."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","criteria":[{"criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"covered","evidence":["src/anomalyzer/recipes.py:344","src/anomalyzer/recipes.py:388","tests/test_analysis.py:441"],"notes":"The injected spike retains its observed value, expectation, standardized residual, and point trigger."},{"criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"covered","evidence":["src/anomalyzer/baseline.py:31","src/anomalyzer/recipes.py:174","tests/test_analysis.py:441","tests/test_trend.py:36"],"notes":"Assertions establish repeatability and unchanged earlier evidence under later evaluation changes with fixed seasonal settings. Automatic positional season discovery has the separately documented streaming limitation."},{"criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"covered","evidence":["src/anomalyzer/evidence.py:95","tests/test_analysis.py:274"],"notes":"The test checks one three-sample run, exact contributing positions, direction, and first detection."},{"criterion":"Separated point departures remain separate episodes, and positional episode intervals use integer coordinates.","status":"covered","evidence":["src/anomalyzer/evidence.py:12","src/anomalyzer/evidence.py:23","tests/test_analysis.py:263","tests/test_analysis.py:369"],"notes":"Separated spikes produce two episodes without a consecutive-run pattern; positional evidence produces interval [9, 9]."}],"ignore_findings":[],"done":true,"input_fingerprint":"062833cc1bd19d5734b0f3c63d1e81666ffb568ead3399e359d62db2c0c0677d","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Find spikes and drops"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","label":"Know when the expected bas...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","acceptance":["Repeating seasonal structure with stable calibration can yield supported readiness.","Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","Insufficient calibration support yields not_assessed, including when the run otherwise completes.","Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment."],"guidance":["Readiness does not prove real-world effectiveness and is not an automatic action-policy gate."],"path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/readiness.py","src/anomalyzer/recipes.py","tests/test_readiness.py"],"automatedChecks":[{"id":"supported-reference","description":"Support repeated seasonality and preserve readiness under later changes.","command":"python -m pytest -q tests/test_readiness.py::test_repeating_season_and_trend_support_reference"},{"id":"weak-season","description":"Warn on unstable seasonality even when no findings occur.","command":"python -m pytest -q tests/test_readiness.py::test_unstable_declared_season_warns_without_calling_series_inapplicable tests/test_readiness.py::test_documented_random_example_has_no_findings_but_warns_on_reference"},{"id":"calibration-drift","description":"Warn when residual calibration drifts despite a repeated training season.","command":"python -m pytest -q tests/test_readiness.py::test_calibration_drift_warns_even_when_training_season_repeats"},{"id":"not-assessed","description":"Distinguish uncalibrated history and too few calibration samples.","command":"python -m pytest -q tests/test_readiness.py::test_readiness_is_not_assessed_without_calibrated_evaluation tests/test_readiness.py::test_short_calibration_is_not_assessed_even_when_analysis_completes"}],"manualChecks":["Read readiness reasons alongside findings; a lack of findings alone does not establish a trustworthy reference."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","criteria":[{"criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"covered","evidence":["src/anomalyzer/readiness.py:128","tests/test_readiness.py:8"],"notes":"Assertions establish repeated seasonality, strong calibration improvement, and supported readiness."},{"criterion":"Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","status":"covered","evidence":["src/anomalyzer/readiness.py:113","src/anomalyzer/readiness.py:122","tests/test_readiness.py:28","tests/test_readiness.py:41"],"notes":"Completed analyses retain caution reasons for weak seasonal repetition or calibration-center drift."},{"criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"covered","evidence":["src/anomalyzer/readiness.py:77","tests/test_readiness.py:56","tests/test_readiness.py:63"],"notes":"Both uncalibrated history and a completed run with three calibration samples return not_assessed."},{"criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"covered","evidence":["src/anomalyzer/recipes.py:327","tests/test_readiness.py:8"],"notes":"Readiness reads training/calibration inputs once; tests compare identical segment and aggregate assessments after a later shift."}],"ignore_findings":[],"done":true,"input_fingerprint":"d01367627e058ea39c78ee0e3616061a2e0725d06e8acd0b84b46a9655431e57","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Assess baseline quality"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","label":"Follow evolving growth whi...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","acceptance":["Each rolling expectation uses only observations before the sample being scored.","Persistent growth changes can be followed more closely than by the frozen default recipe.","An isolated extreme value is held out of future model references rather than creating a seasonal echo.","Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected."],"guidance":["A persistent third extreme miss in one direction may be admitted to the rolling reference; this is not a manual regime reset."],"path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"causal-window","description":"Keep a known rolling expectation invariant when future values are appended.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_expectation_is_one_step_and_prefix_invariant"},{"id":"changing-growth","description":"Follow a persistent growth change and report model admissions.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_tracks_a_persistent_growth_change"},{"id":"isolated-extreme","description":"Exclude an isolated extreme from later reference estimates.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_contains_an_isolated_extreme"},{"id":"bounded-settings","description":"Reject conflicting or unbounded adaptive settings.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_settings_reject_conflicting_or_unbounded_trend"}],"manualChecks":["Inspect the adaptive recipe's rolling-window parameters and frozen calibration before interpreting its final expectation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","criteria":[{"criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"covered","evidence":["src/anomalyzer/baseline.py:61","tests/test_analysis.py:48"],"notes":"All rolling-reference positions precede the scored index. The independent expectation remains unchanged after future values are appended."},{"criterion":"Persistent growth changes can be followed more closely than by the frozen default recipe.","status":"covered","evidence":["src/anomalyzer/recipes.py:145","src/anomalyzer/recipes.py:370","tests/test_analysis.py:58"],"notes":"The test requires adaptive held-out MAE below half the frozen recipe's MAE and a close final expectation."},{"criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","tests/test_analysis.py:81"],"notes":"The isolated spike is excluded from model references; its next seasonal expectation closely matches the control. Persistent third misses are admitted as guidance explicitly permits."},{"criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:41","tests/test_analysis.py:104"],"notes":"Both incompatible explicit trend selection and excessive slope lookback raise specific validation errors."}],"ignore_findings":[],"done":true,"input_fingerprint":"984422f1fd3ab3d22cf7b4ba9d0bd06629fc228a8e53d7ad44439cd8d35ec0db","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Follow evolving growth"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","label":"Account for daily and week...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"planned","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","acceptance":["Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","Insufficient support for a requested cycle produces an explicit limitation or declared fallback rather than a silently fitted component.","Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle."],"guidance":["Begin with a bounded daily-and-weekly use case; overlapping cycles share information and cannot be treated as independent evidence.","Proposed automated checks cover combined ordinary cycles, future isolation, unsupported cycle history, and extreme-value influence across cycles.","Manual review should inspect a weekday morning spike and verify the baseline explains ordinary calendar variation without explaining away the spike."],"path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Represent more than one repeating cycle"},"impact":0,"gaps":3,"r":47,"x":170,"y":406},{"id":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","label":"Limit how extreme values d...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","acceptance":["Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","Training and calibration diagnostics identify excluded extreme observations.","An extreme evaluated sample remains evidence while its expected value is used in future references.","Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior."],"guidance":["Protection keeps the observed evidence; it does not delete observations or certify that a departure was an incident."],"path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/robust.py","src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py","tests/test_cli.py"],"automatedChecks":[{"id":"robust-reference","description":"Limit an extreme weekly reference and use earlier references at lag one.","command":"python -m pytest -q tests/test_analysis.py::test_robust_weekly_reference_limits_one_high_week tests/test_analysis.py::test_robust_lag_one_uses_prior_seasonal_values"},{"id":"training-calibration","description":"Record training and calibration exclusions and retain a useful growth fit.","command":"python -m pytest -q tests/test_trend.py::test_robust_training_fit_and_calibration_exclusions"},{"id":"include-choice","description":"Compare robust and include modes through the CLI and show their differing triggers.","command":"python -m pytest -q tests/test_cli.py::test_outlier_handling_cli"}],"manualChecks":["Inspect excluded_from_model, reference_action, and outlier-handling diagnostics alongside the original observed value."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","criteria":[{"criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"covered","evidence":["src/anomalyzer/baseline.py:39","tests/test_analysis.py:25"],"notes":"Assertions check a independently computed weighted reference and invariance when an already-trimmed high value becomes more extreme."},{"criterion":"Training and calibration diagnostics identify excluded extreme observations.","status":"covered","evidence":["src/anomalyzer/trend.py:178","src/anomalyzer/recipes.py:287","tests/test_trend.py:65"],"notes":"Tests identify exact training/calibration exclusions and compare their influence on trend selection and residual scale."},{"criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","src/anomalyzer/recipes.py:392","tests/test_analysis.py:263","tests/test_trend.py:11"],"notes":"The observed departure and trigger survive; only reference_values receives the prior expectation, preventing the next seasonal echo."},{"criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"covered","evidence":["src/anomalyzer/recipes.py:151","src/anomalyzer/recipes.py:475","tests/test_cli.py:95"],"notes":"CLI assertions distinguish robust triggers {65} from include-mode triggers {65, 72}."}],"ignore_findings":[],"done":true,"input_fingerprint":"183b8696fdc9786878b3e54c4a9e5f68180b47e157528bef953274a3487d49a7","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Protect later comparisons"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","label":"Compare against a new base...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","acceptance":["Integer positions and exact observed dates or timestamps can identify reset boundaries.","Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","Repeated, out-of-range, or unmatched reset points are rejected.","CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries."],"guidance":["Reset points are explicit caller decisions. Known-event notes and review-state fields do not rebuild regimes.","Multi-resolution analysis does not support reset_points."],"path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py"],"automatedChecks":[{"id":"independent-regimes","description":"Retrain separate regimes without cross-boundary evidence or patterns.","command":"python -m pytest -q tests/test_analysis.py::test_manual_resets_retrain_without_crossing_boundaries"},{"id":"invalid-boundaries","description":"Reject invalid or duplicate resolved reset positions.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_manual_resets"},{"id":"cli-boundaries","description":"Resolve position and date reset flags through the CLI.","command":"python -m pytest -q tests/test_cli.py::test_manual_reset_cli"}],"manualChecks":["Inspect the resolved reset positions and segment diagnostics for examples/steps_reset.json."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","criteria":[{"criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"covered","evidence":["src/anomalyzer/validation.py:129","tests/test_analysis.py:381"],"notes":"Positions and observed dates are exercised end to end; timestamp normalization and exact matching are implemented directly."},{"criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"covered","evidence":["src/anomalyzer/recipes.py:119","src/anomalyzer/recipes.py:143","tests/test_analysis.py:381"],"notes":"Each segment initializes separate state and passes its own start to forecasts. Tests verify three calibrated regimes and no cross-boundary evidence or findings."},{"criterion":"Repeated, out-of-range, or unmatched reset points are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/validation.py:149","tests/test_analysis.py:413"],"notes":"Validation rejects literal duplicates and different declarations resolving to the same position, invalid positions, and unmatched dates/timestamps."},{"criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"covered","evidence":["tests/test_cli.py:82","src/anomalyzer/recipes.py:53"],"notes":"CLI assertions establish positions [50, 100] and removal of the unmarked-regime episodes."}],"ignore_findings":[],"done":true,"input_fingerprint":"067063f972c5b70d990d6cb0125892731a97a7f482698d7b331a1a30229ac223","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Restart after a reviewed change"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","label":"Choose the repeating cycle...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","acceptance":["An explicit season length is honored and reported as explicit.","A repeating positional series can infer a seasonal lag and report the selection window.","A positional series without a sufficiently stable candidate falls back to lag one and records why.","The reserved calibration and evaluation tail does not determine positional season selection."],"guidance":["Candidate discovery uses the input prefix before the reserved tail and confirms the candidate in an early window. It is not fully prefix-invariant streaming discovery.","Supply season_length explicitly when stable streaming season selection is required."],"path":".capabilities/analysis/expectations/select-season-length.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/seasonality.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"declared-season","description":"Honor a separate explicit season setting.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"inference-fallback","description":"Infer a repeating lag or record lag-one fallback when no season is supported.","command":"python -m pytest -q tests/test_analysis.py::test_positional_season_inference_and_fallback"},{"id":"reserved-tail","description":"Keep selection unchanged when only the reserved tail changes.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_season_inference_excludes_reserved_tail"},{"id":"cli-inference","description":"Expose the inferred lag and its source for bare JSON values in the CLI.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"}],"manualChecks":["Review selection_window and ranked candidates for a positional series before treating its inferred season as an ongoing monitoring assumption."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","criteria":[{"criterion":"An explicit season length is honored and reported as explicit.","status":"covered","evidence":["src/anomalyzer/recipes.py:55","tests/test_analysis.py:234"],"notes":"Explicit field presence bypasses inference, and the test checks the reported source."},{"criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"covered","evidence":["src/anomalyzer/seasonality.py:96","tests/test_analysis.py:247"],"notes":"The repeating seven-position series selects lag 7 and reports confirmation window [0, 27]."},{"criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"covered","evidence":["src/anomalyzer/seasonality.py:79","src/anomalyzer/seasonality.py:84","tests/test_analysis.py:247"],"notes":"Candidate-strength and early-confirmation failures return lag 1 with a reason; the quadratic example verifies fallback."},{"criterion":"The reserved calibration and evaluation tail does not determine positional season selection.","status":"covered","evidence":["src/anomalyzer/seasonality.py:69","tests/test_capability_behaviors.py:147"],"notes":"Discovery excludes calibration_size plus one reserved sample. Changing that tail preserves the complete inference diagnostics; arbitrary appended-prefix invariance is explicitly outside the guidance."}],"ignore_findings":[],"done":true,"input_fingerprint":"3317b1727dc3cabfcfc9b87808d9249d0d59f111b8fd5f8a1fad653b21343980","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Choose a repeating cycle"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","label":"Interpret rate changes wit...","area":"analysis","scopes":["analysis","analysis/rates"],"scope":"analysis/rates","scopeLabel":"analysis/rates","status":"planned","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","acceptance":["Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","Zero denominators remain unavailable, and low support is explicitly marked; intervals remain within the valid zero-to-one rate domain.","Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations."],"guidance":["Independence and stable exposure are modeling assumptions; clustered events, overdispersion, or changing populations require explicit limitations.","Proposed automated checks cover zero denominators, tiny counts, zero-or-all-success boundaries, and equal rates at different volumes.","Manual review should compare sparse and high-volume conversion changes and confirm the explanation discloses assumptions without inventing incident probability."],"path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Judge rates using their exposure"},"impact":0,"gaps":3,"r":47,"x":194,"y":524},{"id":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","label":"See changes before the day...","area":"analysis","scopes":["analysis","analysis/resolution"],"scope":"analysis/resolution","scopeLabel":"analysis/resolution","status":"implemented","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","acceptance":["Callers can select sum, mean, or last aggregation for completed and partial periods.","Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","An explicitly unfinished source sample is excluded from both analysis views.","A source cadence that does not evenly divide 24 hours is rejected."],"guidance":["Periods are fixed 24-hour windows, not daylight-saving-aware local calendar days.","Native and daily views share source observations and are correlated.","This recipe supports schema 1.0 and does not support reset_points."],"path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","dependencies":["data/quality/exclude-incomplete-periods","analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"selected-aggregation","description":"Check sum, mean, and last for completed days and trailing partial aggregates.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_multi_resolution_uses_selected_aggregation"},{"id":"two-views","description":"Return completed days, native evidence, and a linked trailing partial-day marker.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_current_period_and_keeps_two_views"},{"id":"unfinished-source","description":"Exclude a marked incomplete source sample from both views.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days"},{"id":"dividing-cadence","description":"Reject subperiod durations that do not evenly divide a day.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_rejects_non_dividing_frequency"}],"manualChecks":["Inspect the partial aggregate's observed_aggregate and intraday_evidence_refs together with the two method namespaces."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","criteria":[{"criterion":"Callers can select sum, mean, or last aggregation for completed and partial periods.","status":"covered","evidence":["src/anomalyzer/contracts.py:18","src/anomalyzer/recipes.py:539","tests/test_capability_behaviors.py:178"],"notes":"Parameterized assertions check independently known completed and partial values for all three aggregation choices."},{"criterion":"Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","status":"covered","evidence":["src/anomalyzer/recipes.py:617","src/anomalyzer/recipes.py:667","src/anomalyzer/recipes.py:680","tests/test_analysis.py:170"],"notes":"Separate datasets, configurations, and namespaces preserve the two correlated views."},{"criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"covered","evidence":["src/anomalyzer/recipes.py:639","src/anomalyzer/recipes.py:690","tests/test_analysis.py:170"],"notes":"The marker contains finalized progress, observed aggregate, native evidence references, and an explicit exclusion flag."},{"criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"covered","evidence":["src/anomalyzer/validation.py:44","src/anomalyzer/recipes.py:564","tests/test_analysis.py:200"],"notes":"Both child datasets use prepared finalized observations; tests verify counts and retention of the excluded original value."},{"criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"covered","evidence":["src/anomalyzer/recipes.py:572","tests/test_analysis.py:221"],"notes":"A five-hour cadence raises the specific dividing-cadence error."}],"ignore_findings":[],"done":true,"input_fingerprint":"cdcdea1166a4e3d2523b594bec938142135e17a6b4279732e52819a486d1b4f2","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Compare finalized subperiods"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","label":"Compare days and weeks in ...","area":"data","scopes":["data","data/calendar"],"scope":"data/calendar","scopeLabel":"data/calendar","status":"planned","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","acceptance":["Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","Periods report UTC boundaries and local labels; daylight-saving days can contain 23 or 25 hours without being forced into fixed 24-hour windows.","Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation."],"guidance":["Calendar grouping needs a distinct contract from fixed elapsed cadence; it must not silently reinterpret existing frequency or aggregation settings.","Proposed automated checks cover spring transitions, fall transitions, week boundaries, and missing native slots in a calendar period.","Manual review should inspect a daylight-saving week and confirm local labels, observed coverage, and completed-versus-open comparisons are understandable."],"path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Declare local calendar boundaries"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","label":"Analyze many entities with...","area":"data","scopes":["data","data/cohorts"],"scope":"data/cohorts","scopeLabel":"data/cohorts","status":"planned","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","acceptance":["Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","Training, calibration, relationships, and cases do not cross entity boundaries unless a separate explicit aggregate comparison is requested.","The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident."],"guidance":["Batch orchestration should preserve existing per-request contracts rather than silently lifting schema limits; multiple-testing limits stay visible.","Proposed automated checks cover partition isolation, batch-versus-single equivalence, one-entity failure, and shared resource-budget exhaustion.","Manual review should inspect a mixed batch of healthy, changed, and sparse entities and confirm the summary does not equate missing support with health."],"path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Keep each entity's evidence separate"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","label":"Know when missing updates ...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"planned","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","acceptance":["Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","Results distinguish current, late, missing, and not-assessed freshness and expose the last observed update and the deadline used.","A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision."],"guidance":["Freshness depends on declared collection expectations; it does not prove source completeness and does not start a scheduler or send a notification.","Proposed automated checks cover exact deadlines, allowed lateness, absent source updates, and stale-source propagation to relationships.","Manual review should compare a stale feed with a true metric drop and confirm the report distinguishes collection uncertainty from numerical departure."],"path":".capabilities/data/quality/report-source-freshness.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Check whether the sources are current"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","label":"See how late or corrected ...","area":"data","scopes":["data","data/revisions"],"scope":"data/revisions","scopeLabel":"data/revisions","status":"planned","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","acceptance":["A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","Results preserve prior and new evidence versions and identify findings that were added, changed, or withdrawn with traceable source lineage.","Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results."],"guidance":["A corrected event-time result does not replace the record of what was known earlier; callers control retention and any notification about a revision.","Proposed automated checks cover late completion, corrected calibration data, downstream relationship changes, and conflicting source versions.","Manual review should inspect a drop withdrawn after ingestion completes and confirm both the original limitation and the revision explanation are retained."],"path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Reconcile revisions without losing history"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","label":"Choose settings using held...","area":"evaluation","scopes":["evaluation","evaluation/comparisons"],"scope":"evaluation/comparisons","scopeLabel":"evaluation/comparisons","status":"planned","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","acceptance":["Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","The output preserves every candidate's settings and evaluation provenance; adopting a configuration requires an explicit caller decision."],"guidance":["Bounded comparison is not unrestricted parameter search; held-out results describe the supplied history and do not guarantee future effectiveness.","Proposed automated checks cover split isolation, comparable replay inputs, failed candidate retention, and unchanged active settings after comparison.","Manual review should compare an earlier but noisier configuration with a quieter one and confirm the tradeoff is visible without an unsupported winner."],"path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Review the tradeoff before adopting settings"},"impact":0,"gaps":3,"r":47,"x":560.6666666666667,"y":524},{"id":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","label":"Measure detection timing a...","area":"evaluation","scopes":["evaluation","evaluation/outcomes"],"scope":"evaluation/outcomes","scopeLabel":"evaluation/outcomes","status":"planned","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","acceptance":["Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","Timing reports distinguish the incident's supplied start from the first available candidate, calibrated detection, and policy-eligible evidence.","Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown."],"guidance":["Detection timing does not measure response effectiveness or establish that an unmatched finding was false; incident labels are caller judgments.","Proposed automated checks cover boundary matching, ambiguous overlaps, incomplete label coverage, and separate candidate versus eligibility timing.","Manual review should inspect a matched and an unmatched episode and confirm the reported timing and accuracy limits agree with the supplied label coverage."],"path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Compare detection with reviewed outcomes"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","label":"Replay history as it would...","area":"evaluation","scopes":["evaluation","evaluation/replay"],"scope":"evaluation/replay","scopeLabel":"evaluation/replay","status":"planned","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","acceptance":["Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","Repeating a replay preserves numerical evidence and first-detection timing apart from generated IDs and runtime; stopped replays retain completed steps."],"guidance":["Event-time replay does not reconstruct arrival-time availability unless the caller supplies versioned arrival history.","Proposed automated checks cover future-tail isolation, short-history cutoffs, deterministic repeated replay, and interrupted replay retention.","Manual review should inspect one change from its first available observation through calibration and confirm the explanation uses only then-known evidence."],"path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Reconstruct the evidence available at the time"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","label":"Update evidence when the n...","area":"execution","scopes":["execution","execution/incremental"],"scope":"execution/incremental","scopeLabel":"execution/incremental","status":"planned","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","acceptance":["Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","Supported explicitly configured recipes produce the same numerical evidence as full batch analysis of the combined history, including cross-chunk patterns.","Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations."],"guidance":["Start with explicit season lengths; input-prefix season discovery is not currently guaranteed to be invariant when the history grows.","Proposed automated checks cover chunk-versus-batch equivalence, cross-chunk CUSUM, idempotent retry, and failed-append state preservation.","Manual review should follow a sustained change across several append batches and confirm detection timing and provisional evidence remain understandable."],"path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Analyze the next finalized samples"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","label":"Keep useful evidence when ...","area":"execution","scopes":["execution","execution/resilience"],"scope":"execution/resilience","scopeLabel":"execution/resilience","status":"implemented","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","acceptance":["A method failure preserves evidence completed before the failure and reports the error.","Runtime budgets are checked cooperatively and produce an explicit stop reason.","CLI byte limits and execution failures produce distinct error exits.","A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial."],"guidance":["Default limits are 10,000 source observations per request, 10 MB CLI input, and 60 seconds cooperative runtime.","Numerical results are deterministic for fixed declared settings; generated run IDs and runtimes vary."],"path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/budget.py","src/anomalyzer/recipes.py","src/anomalyzer/orchestrator.py","src/anomalyzer/io.py","src/anomalyzer/cli.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"method-and-budget","description":"Retain pre-failure evidence and report a runtime-budget stop.","command":"python -m pytest -q tests/test_analysis.py::test_failure_and_budget"},{"id":"cli-limits","description":"Enforce byte limits and distinguish execution failure in CLI output and exit status.","command":"python -m pytest -q tests/test_cli.py::test_size_limit_and_execution_exit"},{"id":"partial-siblings","description":"Preserve healthy analysis when a sibling dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Inspect run status, stop_reason, and each target's status before relying on a result; confirm limits are presented as cooperative rather than hard process timeouts."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","criteria":[{"criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"covered","evidence":["src/anomalyzer/recipes.py:400","src/anomalyzer/recipes.py:448","tests/test_analysis.py:509"],"notes":"The failure regression preserves evidence through index 44 and reports the injected error at index 45."},{"criterion":"Runtime budgets are checked cooperatively and produce an explicit stop reason.","status":"covered","evidence":["src/anomalyzer/budget.py:18","src/anomalyzer/recipes.py:108","src/anomalyzer/recipes.py:508","tests/test_analysis.py:509"],"notes":"Cooperative target/sample checks expose runtime_budget and budget_exceeded; this is not a hard process timeout."},{"criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:429","tests/test_cli.py:108"],"notes":"The full suite executed the byte-limit and execution-exit regression successfully; invalid input exits 2 and execution failure exits 1."},{"criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:125","src/anomalyzer/orchestrator.py:171","src/anomalyzer/orchestrator.py:185","tests/test_capability_behaviors.py:159","tests/test_capability_behaviors.py:30"],"notes":"Healthy dataset evidence survives target-specific failure; inspected incomplete-usage reports partial with no invented healthy relationship."}],"ignore_findings":[],"done":true,"input_fingerprint":"146a9c8396fd7a2d039f75077dc0b82a0778685d20c3782aa284d5f0b6ed94be","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect incomplete execution"},"impact":0,"gaps":0,"r":38,"x":720,"y":642},{"id":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","label":"Resume local analysis afte...","area":"execution","scopes":["execution","execution/state"],"scope":"execution/state","scopeLabel":"execution/state","status":"planned","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","acceptance":["A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","Restoring a supported state and continuing yields the same numerical evidence and detection positions as uninterrupted processing of the same observations.","Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled."],"guidance":["Stored observations and context can contain sensitive data; document exactly what a snapshot contains and avoid automatic persistence.","Proposed automated checks cover restart equivalence, corrupt snapshots, version or settings mismatch, and output overwrite protection.","Manual review should inspect a snapshot and resumed explanation and confirm their provenance and retained-data scope are clear to an application owner."],"path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Preserve state across restarts"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","label":"Understand findings in a r...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","acceptance":["Point-anomaly episode counts and residual-pattern findings are labeled separately.","Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","Readiness status remains visible, including when it was not assessed."],"guidance":["Readable summaries should retain the evidence's claim limits and allow drilldown into JSON."],"path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/cli.py","tests/test_cli.py","examples/agentic/conversion.json","examples/agentic/unit-cost.json"],"automatedChecks":[{"id":"episodes-patterns","description":"Distinguish shift patterns from point episodes in readable output.","command":"python -m pytest -q tests/test_cli.py::test_readable_output_distinguishes_episodes_from_patterns"},{"id":"relationship-case","description":"Explain a conversion case with entity, units, comparison, and a suggested check.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_relationship_request_through_cli"},{"id":"explicit-rule","description":"Describe the unit-cost rule boundary and unassessed readiness.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the conversion and unit-cost example output and confirm the descriptions state the supported numerical comparison in plain language."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","criteria":[{"criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"covered","evidence":["src/anomalyzer/cli.py:347","src/anomalyzer/cli.py:405","tests/test_cli.py:143"],"notes":"Readable tests distinguish zero-episode residual shifts from a spike episode and retain the overlap warning."},{"criterion":"Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","status":"covered","evidence":["src/anomalyzer/cli.py:291","src/anomalyzer/cli.py:308","src/anomalyzer/cli.py:331","tests/test_cli.py:117","examples/agentic/conversion.json"],"notes":"Conversion output identifies site=demo, observed/expected orders_per_visit, the relationship, and a collection-completeness investigation question."},{"criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"covered","evidence":["src/anomalyzer/cli.py:318","tests/test_cli.py:134","examples/agentic/unit-cost.json"],"notes":"Unit-cost output names review-above-13c and its [0, 0.13] boundary as a declared rule crossing."},{"criterion":"Readiness status remains visible, including when it was not assessed.","status":"covered","evidence":["src/anomalyzer/cli.py:352","src/anomalyzer/cli.py:399","tests/test_cli.py:134"],"notes":"Readable output includes not_assessed and its supporting reasons without claiming measured alert accuracy."}],"ignore_findings":[],"done":true,"input_fingerprint":"98108868b000972fd9b3c3f082846fbd3f3b08c399ef7ed424c23245d6935fef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Read the findings"},"impact":0,"gaps":0,"r":38,"x":903.3333333333334,"y":642},{"id":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","label":"Share an investigation tha...","area":"interfaces","scopes":["interfaces","interfaces/export"],"scope":"interfaces/export","scopeLabel":"interfaces/export","status":"planned","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","acceptance":["Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","Source observations and context are included only by explicit selection; omitted data and resulting limits on independent reproduction are disclosed."],"guidance":["Fingerprints identify included provenance; they do not prove source truth. A portable report must make privacy scope and reproduction limits inspectable.","Proposed automated checks cover manifest consistency, partial-run export, omitted-source disclosure, and overwrite protection.","Manual review should open a bundle outside the original session and confirm a reviewer can trace one finding and see exactly what cannot be reproduced."],"path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Preserve a portable evidence record"},"impact":0,"gaps":3,"r":47,"x":903.3333333333334,"y":642},{"id":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","label":"Install a versioned analys...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","acceptance":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"guidance":["Guard owns scheduling, persistence, business severity, incidents and notifications.","A published package does not establish integration into UsageTap Guard."],"path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","interfaces/cli/analyze-files-and-streams","interfaces/mcp/connect-read-only-local-server"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["pyproject.toml","src/anomalyzer/_version.py","scripts/release.py","scripts/publish_release.py","scripts/verify_distribution.py","scripts/verify_lambda.sh","examples/library/guard_smoke.py",".github/workflows/ci.yml",".github/workflows/release.yml","docs/releases/README.md","tests/test_release.py"],"automatedChecks":[{"id":"release-safety","description":"Reject mismatched tags, altered assets and mutable publication.","command":"python -m pytest -q tests/test_release.py"},{"id":"installed-artifacts","description":"Install wheel and sdist outside the checkout and invoke public APIs.","command":"python scripts/verify_distribution.py --python 3.12"},{"id":"release-inventory","description":"Check archive contents, version metadata and SHA-256 inventory.","command":"python scripts/release.py verify"},{"id":"linux-offline-consumer","description":"Bundle and run with networking disabled in Linux Python 3.12.","command":"bash scripts/verify_lambda.sh"}],"manualChecks":["Follow the release guide and confirm Guard retains domain severity and incident ownership; inspect readiness and evidence limits before adopting."],"verificationGaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for interfaces/python/install-versioned-library.","criteria":[{"criterion":"Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","status":"partial","evidence":["src/anomalyzer/_version.py:1","scripts/release.py:1","scripts/release.py:20","scripts/release.py:39","scripts/release.py:48","scripts/release.py:54","scripts/publish_release.py:19","scripts/publish_release.py:27","scripts/publish_release.py:29","scripts/publish_release.py:39","scripts/verify_lambda.sh:14","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19","scripts/verify_lambda.sh:25","scripts/verify_lambda.sh:27",".github/workflows/ci.yml:14",".github/workflows/ci.yml:19",".github/workflows/ci.yml:27",".github/workflows/ci.yml:41",".github/workflows/ci.yml:46",".github/workflows/release.yml:28","docs/releases/README.md:12","docs/releases/README.md:13","docs/releases/README.md:14","docs/releases/README.md:17","docs/releases/README.md:18","tests/test_release.py:60","tests/test_release.py:71","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","status":"partial","evidence":["scripts/verify_distribution.py:1","scripts/verify_distribution.py:42","examples/library/guard_smoke.py:32","examples/library/guard_smoke.py:49","examples/library/guard_smoke.py:65","examples/library/guard_smoke.py:69","examples/library/guard_smoke.py:75",".github/workflows/ci.yml:57","docs/releases/README.md:13","docs/releases/README.md:208","docs/releases/README.md:211","docs/releases/README.md:213"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:34","scripts/release.py:81","scripts/release.py:82","scripts/release.py:85","scripts/release.py:92","scripts/publish_release.py:1","scripts/publish_release.py:25","scripts/publish_release.py:26","scripts/publish_release.py:27","scripts/publish_release.py:28","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:24","examples/library/guard_smoke.py:39",".github/workflows/ci.yml:42",".github/workflows/ci.yml:45",".github/workflows/ci.yml:46",".github/workflows/ci.yml:49",".github/workflows/ci.yml:51",".github/workflows/release.yml:31",".github/workflows/release.yml:33",".github/workflows/release.yml:36",".github/workflows/release.yml:37","docs/releases/README.md:3","docs/releases/README.md:12","docs/releases/README.md:14","docs/releases/README.md:15","docs/releases/README.md:19","tests/test_release.py:1","tests/test_release.py:21","tests/test_release.py:23","tests/test_release.py:52","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:81","scripts/release.py:82","scripts/publish_release.py:40","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:6","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19",".github/workflows/ci.yml:30",".github/workflows/ci.yml:31",".github/workflows/ci.yml:51",".github/workflows/ci.yml:52",".github/workflows/ci.yml:90","docs/releases/README.md:12","docs/releases/README.md:15","docs/releases/README.md:58","docs/releases/README.md:59","docs/releases/README.md:75"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","partial: Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","partial: Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","partial: A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","name":"Python integration","label":"Python integration / Embed analysis in my application / Install a released dependency"},"impact":0,"gaps":1,"r":41,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","label":"Ask an agent to explain un...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","acceptance":["The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit."],"guidance":["This contract covers the supplied instructions and demonstrated analysis workflow; it does not guarantee that every agent host follows the instructions.","Synthetic examples demonstrate mechanics, not real-world detection accuracy."],"path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/agents/openai.yaml","src/anomalyzer/cli.py","src/anomalyzer/mcp.py","tests/test_cli.py","tests/test_agentic.py","tests/test_mcp.py"],"automatedChecks":[{"id":"structured-cli","description":"Exercise the documented structured CLI workflow and preserve incomplete-history evidence.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence tests/test_cli.py::test_bare_json_values_cli"},{"id":"declared-relationships","description":"Check explicit ratio semantics and unavailable source evidence.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"mcp-drilldown","description":"Follow compact analysis to a finding and its source evidence.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case"}],"manualChecks":["Follow the conversion example through the CLI and MCP drilldown; compare the Skill instructions with actual evidence and check that its proposed text separates calibrated findings, early candidates, readiness, and caller-owned actions."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","criteria":[{"criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:13",".agents/skills/anomaly-detection/SKILL.md:41","tests/test_cli.py:13"],"notes":"Executable resolution and JSON commands are explicit. Manual conversion and spike CLI runs use the installed numerical engine."},{"criterion":"The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:24",".agents/skills/anomaly-detection/SKILL.md:32","tests/test_cli.py:51","tests/test_agentic.py:211"],"notes":"Instructions preserve scope, units, missingness, declared relationships, and exact alignment; existing runtime checks establish these prerequisites."},{"criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:101",".agents/skills/anomaly-detection/SKILL.md:134",".agents/skills/anomaly-detection/agents/openai.yaml:4"],"notes":"The supplied instructions lead with plain-language changes and then qualify maturity/readiness/overlap. This establishes instructional content, not measured agent compliance."},{"criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:54",".agents/skills/anomaly-detection/SKILL.md:174","src/anomalyzer/mcp.py:169","tests/test_mcp.py:54"],"notes":"The documented three-tool workflow matches executed conversion drilldown, with session-only run IDs and explicit read-only/caller-owned delivery limits."}],"ignore_findings":[],"done":true,"input_fingerprint":"751723214aa340b1227ec347c22f130749fcdbfd49327c21b4c27ef68d7f0341","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Get an evidence-grounded explanation"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","label":"See the evidence in a char...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","acceptance":["The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes."],"guidance":["Native chart rendering and interactive figure features depend on the agent host; the checked-in fallback produces a static SVG for a single dataset.","A chart explains structured analysis; it does not replace the detector result."],"path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/scripts/quick_chart.py",".agents/skills/anomaly-detection/references/visualization.md","tests/test_skill_chart.py"],"automatedChecks":[{"id":"observed-evidence","description":"Render real observed and expected data and only actual point triggers.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_preserves_real_point_evidence"},{"id":"maturity-and-resets","description":"Keep provisional evidence non-triggering and label reviewed regimes.","command":"python -m pytest -q tests/test_skill_chart.py::test_provisional_scores_are_not_charted_as_triggers tests/test_skill_chart.py::test_quick_chart_marks_reviewed_resets"},{"id":"unsupported-scope","description":"Reject a multi-dataset result rather than mixing units.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_rejects_multiple_datasets"},{"id":"output-protection","description":"Require an explicit overwrite for existing charts.","command":"python -m pytest -q tests/test_skill_chart.py::test_chart_cli_requires_explicit_overwrite"}],"manualChecks":["Inspect a spike fallback chart and the detailed figure instructions; verify that axes, initialization context, trigger markers, and interpretation limits are clear, and that the Skill requests a chart only when the user asks."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","criteria":[{"criterion":"The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:189",".agents/skills/anomaly-detection/SKILL.md:226"],"notes":"The Skill explicitly requires a visual request after analysis and separates quick versus detailed rendering instructions."},{"criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:81",".agents/skills/anomaly-detection/scripts/quick_chart.py:59","tests/test_skill_chart.py:28","tests/test_skill_chart.py:43","tests/test_skill_chart.py:54"],"notes":"Real-result tests establish observed/expected lines, trigger-only markers, provisional non-triggering, units, initialization shading, and reset boundaries; the spike chart was inspected visually."},{"criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:85",".agents/skills/anomaly-detection/scripts/quick_chart.py:251","tests/test_skill_chart.py:63","tests/test_skill_chart.py:68"],"notes":"Multiple resolved datasets raise a specific error. The CLI test preserves an existing chart and replaces it only with explicit overwrite."},{"criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/references/visualization.md:31",".agents/skills/anomaly-detection/references/visualization.md:37",".agents/skills/anomaly-detection/references/visualization.md:45",".agents/skills/anomaly-detection/references/visualization.md:86"],"notes":"Detailed instructions require index alignment, maturity distinctions, first detection timing, visible unavailable status, and qualified causal claims; native interactive rendering remains host-dependent."}],"ignore_findings":[],"done":true,"input_fingerprint":"2780b6c3e363f89a30a0709772c0a46a9a96f1567fabb27e63934a3ca80dc982","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Inspect a requested chart"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","label":"Find which segments contri...","area":"investigation","scopes":["investigation","investigation/cohorts"],"scope":"investigation/cohorts","scopeLabel":"investigation/cohorts","status":"planned","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","acceptance":["Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","Segment ranking retains absolute signed contributions and baseline support; an arithmetic contributor is not automatically labeled anomalous or causal."],"guidance":["Ratios and overlapping customer groups are not additive totals; separate segment expectations must not be assumed to sum to a fitted total expectation.","Proposed automated checks cover exact reconciliation, offsetting segment changes, missing segments, and zero total departure.","Manual review should inspect a regional decline with an offsetting increase and confirm the explanation preserves signed contributions and evidence limits."],"path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Locate contributors to a total change"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","label":"Keep reviewed outcomes bes...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","acceptance":["A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","Review export distinguishes analyst judgment from detector output; annotation alone does not change calibration, eligibility, or regime boundaries."],"guidance":["Caller-supplied reviewer identity is attribution, not built-in authentication; storage, access control, response execution, and incident ownership stay with callers.","Proposed automated checks cover evidence-version linking, unresolved reviews, revision retention, and unchanged numerical evidence after annotation.","Manual review should inspect a disputed case with two review revisions and confirm judgments, uncertainties, and original findings remain distinguishable."],"path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Record the reviewed outcome"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","label":"Find earlier reviewed case...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","acceptance":["Search is bounded to caller-selected records and explicit compatibility filters; it does not scan unrelated local files or require a model API.","Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match."],"guidance":["Similar historical evidence does not establish the same cause or authorize repeating a past response; caller-selected records may be incomplete or biased.","Proposed automated checks cover compatibility filtering, bounded corpus access, missing comparison context, and no-match behavior.","Manual review should compare superficially similar cases with different outcomes and confirm precedent is presented as investigation context."],"path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Inspect comparable reviewed history"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","label":"Change action eligibility ...","area":"investigation","scopes":["investigation","investigation/policy"],"scope":"investigation/policy","scopeLabel":"investigation/policy","status":"implemented","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","acceptance":["Default policy leaves action and notification eligibility disabled.","Explicit-rule and calibrated-departure eligibility are controlled independently.","Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged."],"guidance":["Eligibility does not send notifications, execute actions, check readiness, or confer authority to act.","Monitoring schedules, delivery, reviewed incident state, and deduplication belong to the caller."],"path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","dependencies":["investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/agent.py","src/anomalyzer/cases.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"independent-notifications","description":"Replay notification and action choices without changing numerical evidence or the original result.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_separates_action_from_notification"},{"id":"calibrated-choice","description":"Enable calibrated departures only with the corresponding policy choice.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_enables_calibrated_departures_only_when_requested"},{"id":"nontriggering-candidates","description":"Keep descriptive candidates ineligible even under a permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"}],"manualChecks":["Inspect the unit-cost example under default and allow_explicit_rules policies and compare eligibility with the unchanged numerical violation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","criteria":[{"criterion":"Default policy leaves action and notification eligibility disabled.","status":"covered","evidence":["src/anomalyzer/contracts.py:329","tests/test_capability_behaviors.py:74"],"notes":"Default flags disable both eligibility types, including after replay resets policy."},{"criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"covered","evidence":["src/anomalyzer/agent.py:47","tests/test_capability_behaviors.py:123"],"notes":"Explicit-only permission does not enable a calibrated departure; the matching calibrated choice does."},{"criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"covered","evidence":["src/anomalyzer/agent.py:53","tests/test_capability_behaviors.py:101","tests/test_capability_behaviors.py:134"],"notes":"Notification-only replay enables a met rule while keeping action disabled; permissive settings still cannot enable candidates."},{"criterion":"Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged.","status":"covered","evidence":["src/anomalyzer/agent.py:42","src/anomalyzer/cases.py:57","src/anomalyzer/cases.py:105","tests/test_capability_behaviors.py:101","tests/test_agentic.py:49"],"notes":"Assertions compare unchanged original serialization, methods, evidence references, stable IDs, and eligibility-dependent revision changes."}],"ignore_findings":[],"done":true,"input_fingerprint":"ee67bcc1ec1380b8d84a0556d838d648ae6cef3f4f9b9e34e1b71e90eb00aeae","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Review action eligibility"},"impact":0,"gaps":0,"r":38,"x":1110.6666666666667,"y":760},{"id":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","label":"Review a possible new base...","area":"investigation","scopes":["investigation","investigation/regimes"],"scope":"investigation/regimes","scopeLabel":"investigation/regimes","status":"planned","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","acceptance":["A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","Producing a proposal leaves the active regime and original findings unchanged.","A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect."],"guidance":["A proposed boundary is not a confirmed cause or a retrospectively available detection; adaptive reference updates remain a separate behavior.","Proposed automated checks cover unchanged active results, short new regimes, rejected proposals, and accepted explicit-reset equivalence.","Manual review should inspect a persistent outage-like drop and a known launch and confirm neither becomes ordinary behavior merely because it persists."],"path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Review a proposed regime transition"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","label":"Distinguish changing perfo...","area":"relationships","scopes":["relationships","relationships/cohorts"],"scope":"relationships/cohorts","scopeLabel":"relationships/cohorts","status":"planned","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","acceptance":["Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","The report preserves aggregate and cohort rates, population weights, and the declared decomposition formula and reference-period convention.","Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes."],"guidance":["Decomposition is descriptive accounting; it does not estimate an intervention effect or establish that the comparison populations are causally equivalent.","Proposed automated checks cover stable rates with changing mix, stable mix with changing rates, offsetting cohort rates, and new or empty cohorts.","Manual review should inspect a Simpson's-paradox example and confirm the aggregate movement and cohort evidence can both be understood correctly."],"path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Separate performance changes from population mix"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878},{"id":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","label":"Find changes in conversion...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","acceptance":["Ratios retain the computed value, formula, and original source indexes for each sample.","Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","Derived series can be analyzed with the frozen or adaptive single-resolution recipe."],"guidance":["Relationships are caller-supplied declarations, not discovered or causal relationships.","A difference requires identical declared units; a normalized residual requires a positive supplied scale.","A joint condition is a declared 0/1 comparison, not statistical proof."],"path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","dependencies":["relationships/quality/require-compatible-sources","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_examples.py"],"automatedChecks":[{"id":"ratio-lineage","description":"Explain a changing rate with the exact source indexes and formula.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"other-comparisons","description":"Check declared differences, normalized residuals, and combined conditions.","command":"python -m pytest -q tests/test_agentic.py::test_remaining_relationship_primitives"},{"id":"unavailable-values","description":"Keep zero-denominator and missing-source observations unavailable.","command":"python -m pytest -q tests/test_agentic.py::test_missing_and_zero_denominator_are_unavailable_not_zero"},{"id":"adaptive-relationship","description":"Use the adaptive recipe for a declared relationship.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_recipe_runs_for_declared_relationships"}],"manualChecks":["Inspect examples/agentic/conversion.json and distinguish its relationship departure from the individual source metrics."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","criteria":[{"criterion":"Ratios retain the computed value, formula, and original source indexes for each sample.","status":"covered","evidence":["src/anomalyzer/relationships.py:84","src/anomalyzer/relationships.py:152","tests/test_agentic.py:211"],"notes":"Assertions check exact ratio values, formula, and both original source indexes."},{"criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"covered","evidence":["src/anomalyzer/relationships.py:97","src/anomalyzer/relationships.py:134","tests/test_agentic.py:285"],"notes":"Parameterized assertions independently specify difference, normalized residual, all-condition, and any-condition outputs."},{"criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"covered","evidence":["src/anomalyzer/relationships.py:89","src/anomalyzer/relationships.py:142","tests/test_agentic.py:226"],"notes":"Lineage retains null values with missing_source_value/zero_denominator reasons. Explicit zero_denominator=error intentionally fails instead, as documented."},{"criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:87","src/anomalyzer/orchestrator.py:135","tests/test_agentic.py:171","tests/test_examples.py:17"],"notes":"Both supported recipes dispatch derived datasets through the shared single-resolution engine."}],"ignore_findings":[],"done":true,"input_fingerprint":"7fa8f7ffdc0d852ad961b3e1753761d1bd3b00b87b8c001a957ec19e26d068b9","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Compare ratios and metric gaps"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","label":"Check whether a metric fol...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","acceptance":["A supplied lag and coefficient use the predictor at the declared earlier sample position.","Derived values retain response and predictor source indexes and source cutoffs in lineage.","When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","Samples lacking the earlier predictor or awaiting coefficient training remain unavailable."],"guidance":["Lag is measured in samples and is never searched automatically.","Exactly one of coefficient or training_size is required."],"path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","dependencies":["relationships/quality/require-compatible-sources"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_examples.py","examples/agentic/lagged-response.json"],"automatedChecks":[{"id":"supplied-response","description":"Use the declared lag and coefficient with correct predictor indexes.","command":"python -m pytest -q tests/test_agentic.py::test_lagged_response_uses_supplied_coefficient_and_prior_sample"},{"id":"frozen-coefficient","description":"Keep fitted coefficient and earlier lineage invariant when future samples arrive.","command":"python -m pytest -q tests/test_agentic.py::test_estimated_lag_coefficient_is_frozen_and_prefix_invariant"},{"id":"delayed-example","description":"Run the checked-in lagged-response demonstration against expected results.","command":"python -m pytest -q tests/test_examples.py::test_agentic_demo_checks_all_cases"},{"id":"pending-training","description":"Retain unavailable lineage until the full declared training prefix arrives, then freeze its coefficient.","command":"python -m pytest -q tests/test_agentic.py::test_lag_fit_waits_for_complete_training_prefix"}],"manualChecks":["Inspect the lagged-response example's leading unavailable values, coefficient parameters, and first usable derived sample."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","criteria":[{"criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"covered","evidence":["src/anomalyzer/relationships.py:115","tests/test_agentic.py:254"],"notes":"Supplied lag/coefficient uses the declared earlier predictor position and preserves the expected response residual."},{"criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"covered","evidence":["src/anomalyzer/relationships.py:152","tests/test_agentic.py:254","examples/agentic/lagged-response.json"],"notes":"Available and pending lineage retains existing source indexes and timestamp cutoffs; missing predictors do not become zero."},{"criterion":"When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","status":"covered","evidence":["src/anomalyzer/relationships.py:47","tests/test_agentic.py:294","tests/test_agentic.py:330"],"notes":"The coefficient is fitted once after the complete declared prefix; the independent 20/7 result and prefix-invariance assertions pass."},{"criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"covered","evidence":["src/anomalyzer/relationships.py:122","tests/test_agentic.py:330"],"notes":"Short prefixes retain null derived values and explicit unavailable reasons, while an unidentifiable completed fit has a visible error."}],"ignore_findings":[],"done":true,"input_fingerprint":"fc4516ef6ab903dfa4918268d97049aff04fbf9556620ccb8259ff830ba85785","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Check delayed responses"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","label":"Inspect related metrics be...","area":"relationships","scopes":["relationships","relationships/resolution"],"scope":"relationships/resolution","scopeLabel":"relationships/resolution","status":"planned","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed calendar or fixed-duration aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","acceptance":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"guidance":["Initial support should use explicit relationships and compatible boundaries; cross-cadence joining or automatic resampling needs a separate contract.","Proposed automated checks cover ratio-of-totals arithmetic, missing sources, mismatched period boundaries, and open-period evidence separation.","Manual review should inspect a conversion decline within an unfinished day and verify the explanation separates native evidence from partial totals."],"path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/resolution/compare-finalized-subperiods has no automated or manual verification checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/resolution/compare-finalized-subperiods has no automated checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-manual-review","message":"relationships/resolution/compare-finalized-subperiods has no manual review guidance","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Compare related metrics while the period is open"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878}],"links":[{"source":"data/inputs/accept-analysis-requests","target":"analysis/evidence/calibrate-departures","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/cli/analyze-files-and-streams","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"analysis/expectations/model-seasonal-level-and-growth","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"interfaces/python/analyze-through-public-api","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"investigation/assessments/describe-early-changes","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"relationships/quality/require-compatible-sources","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/identify-residual-patterns","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"data/quality/exclude-incomplete-periods","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/mcp/connect-read-only-local-server","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/rules/check-declared-numerical-rules","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/report-point-anomaly-episodes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/evidence/assess-reference-readiness","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/adapt-to-changing-growth","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/protect-reference-from-extremes","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"data/quality/exclude-incomplete-periods","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"execution/resilience/retain-evidence-under-failure-and-limits","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/mcp/connect-read-only-local-server","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/visualize-analysis-on-request","kind":"committed"},{"source":"investigation/rules/check-declared-numerical-rules","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/evaluate-delayed-response","kind":"committed"}]};
+const graph = {"nodes":[{"id":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","label":"Understand how much histor...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","acceptance":["Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","Provisional residual scores can be reported but do not trigger point-anomaly episodes.","Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger."],"guidance":["Evidence strength is not a probability of an incident.","Calibration and expectations are separate concerns; each declared regime calibrates again."],"path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/detection/identify-residual-patterns","analysis/detection/report-point-anomaly-episodes","analysis/evidence/assess-reference-readiness","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","relationships/comparisons/compare-declared-metrics"],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"short-history","description":"Keep reference-only evidence for a short series.","command":"python -m pytest -q tests/test_analysis.py::test_short_series_returns_reference_evidence"},{"id":"provisional-evidence","description":"Report provisional scores without point triggers.","command":"python -m pytest -q tests/test_analysis.py::test_early_evidence_maturity_and_no_provisional_trigger"},{"id":"independent-calibration","description":"Check a known residual center, sample scale, and standardized score.","command":"python -m pytest -q tests/test_analysis.py::test_calibration_reference"},{"id":"constant-reference","description":"Handle zero residual variance without spurious episodes.","command":"python -m pytest -q tests/test_analysis.py::test_constant_and_schema"}],"manualChecks":["Inspect a sample's signal_maturity, calibration_samples, standardized_residual, and triggers together before describing a statistical departure."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","criteria":[{"criterion":"Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","status":"covered","evidence":["src/anomalyzer/recipes.py:205","src/anomalyzer/recipes.py:445","tests/test_analysis.py:141"],"notes":"The short-series test checks three reference-only samples, absent standardized scores, and no triggers."},{"criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"covered","evidence":["src/anomalyzer/recipes.py:210","tests/test_analysis.py:123"],"notes":"A provisional score of 2 exceeds the configured point threshold of 1 but produces no trigger or episode."},{"criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"covered","evidence":["src/anomalyzer/recipes.py:231","src/anomalyzer/recipes.py:307","tests/test_analysis.py:111"],"notes":"Calibration uses retained residual mean and sample standard deviation once per regime; the independent numerical example produces the expected standardized score."},{"criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"covered","evidence":["src/anomalyzer/contracts.py:104","src/anomalyzer/recipes.py:350","src/anomalyzer/recipes.py:388","tests/test_analysis.py:111"],"notes":"Sample fields retain the comparison and support; the point threshold is inspectable in the enclosing method parameters."}],"ignore_findings":[],"done":true,"input_fingerprint":"ccdc1b2e92f3e647d5a9e6127cc3c4b10bf0da8f1f78af3efb805daa9b1e4305","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Establish evidence support"},"impact":6,"gaps":0,"r":68,"x":170,"y":170},{"id":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","label":"Bring my metrics into an a...","area":"data","scopes":["data","data/inputs"],"scope":"data/inputs","scopeLabel":"data/inputs","status":"implemented","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","acceptance":["Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","Published schema 1.0 and 1.1 request and result schemas match the public Python contracts."],"guidance":["Null observations remain unknown unless a schema-1.1 additive-volume policy explicitly authorizes confirmed empty collection as zero. Limits apply to the request as a whole.","Describe current supported behavior; historical plans do not add accepted inputs."],"path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["analysis/evidence/calibrate-departures","analysis/expectations/model-seasonal-level-and-growth","data/quality/analyze-sparse-observations","execution/resilience/retain-evidence-under-failure-and-limits","interfaces/python/analyze-through-public-api","investigation/assessments/describe-early-changes"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_agentic.py","schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json"],"automatedChecks":[{"id":"input-forms","description":"Analyze a bare list and validate a canonical request and result.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings tests/test_analysis.py::test_constant_and_schema"},{"id":"explicit-scope","description":"Reject schema 1.1 datasets without entity scope.","command":"python -m pytest -q tests/test_agentic.py::test_schema_11_requires_explicit_entity_scope"},{"id":"invalid-inputs","description":"Reject unsupported settings and nonnumeric or nonfinite observations.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_config tests/test_analysis.py::test_invalid_values"},{"id":"published-schemas","description":"Compare all four checked-in schemas with their public contracts.","command":"python -m pytest -q tests/test_analysis.py::test_checked_in_schemas tests/test_agentic.py::test_v11_checked_in_schemas_exist"}],"manualChecks":["Review the input examples in README.md and confirm they describe the two supported schema versions and the four-dataset/four-relationship limit."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","criteria":[{"criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"covered","evidence":["src/anomalyzer/recipes.py:21","src/anomalyzer/contracts.py:96","tests/test_analysis.py:234"],"notes":"Bare-list settings are validated; schema 1.0 limits datasets to exactly one."},{"criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"covered","evidence":["src/anomalyzer/contracts.py:240","src/anomalyzer/contracts.py:335","tests/test_agentic.py:192"],"notes":"Entity scope is required, relationship kinds are discriminated, and references must name distinct known datasets."},{"criterion":"Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:7","src/anomalyzer/contracts.py:71","tests/test_analysis.py:526","tests/test_analysis.py:531"],"notes":"Assertions reject each listed category; missing null values remain a separate accepted representation."},{"criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"covered","evidence":["schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json","tests/test_analysis.py:547","tests/test_agentic.py:317"],"notes":"Tests and independent inspection compare each complete published schema with its contract-generated schema. Python cross-field validators supplement JSON Schema constraints."}],"ignore_findings":[],"done":true,"input_fingerprint":"3546cd8f436812ff11c9352a855926677b55cc6287e40f6a5041bd613173e5b4","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Supply a numeric series"},"impact":6,"gaps":0,"r":68,"x":353.33333333333337,"y":170},{"id":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","label":"Analyze local files from t...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","acceptance":["The methods command lists supported analysis recipes and residual detectors.","Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status."],"guidance":["CSV callers must select one entity and unit before analysis.","Exit 0 includes completed analysis with anomalies, insufficient history, or inapplicability; exit 2 means invalid input, and exit 1 means execution failure or partial execution."],"path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/python/install-versioned-library","interfaces/skill/guide-anomaly-investigation","interfaces/skill/visualize-analysis-on-request"],"implementationReferences":["src/anomalyzer/cli.py","src/anomalyzer/io.py","src/anomalyzer/__main__.py","tests/test_cli.py"],"automatedChecks":[{"id":"file-stdin","description":"Analyze equivalent file and stdin requests.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence"},{"id":"csv-precedence","description":"Validate selected CSV fields and apply flag-over-file-over-embedded precedence.","command":"python -m pytest -q tests/test_cli.py::test_csv_and_precedence"},{"id":"value-list","description":"Analyze a bare JSON array and expose inferred settings.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"},{"id":"output-and-exits","description":"Protect existing output and distinguish invalid options from failed execution.","command":"python -m pytest -q tests/test_cli.py::test_output_protection tests/test_cli.py::test_methods_and_invalid_options tests/test_cli.py::test_size_limit_and_execution_exit"}],"manualChecks":["Run an example in readable and JSON modes and confirm stdout contains results while diagnostics go to stderr."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","criteria":[{"criterion":"The methods command lists supported analysis recipes and residual detectors.","status":"covered","evidence":["src/anomalyzer/cli.py:426","src/anomalyzer/registry.py:15","tests/test_cli.py:27"],"notes":"Discovery tests check all three recipes and ten residual detectors, with structured stdout and no successful-run diagnostics on stderr."},{"criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:239","tests/test_cli.py:13","tests/test_mcp.py:163"],"notes":"The passing file/stdin regression compares request fingerprints and short-history semantics; cross-interface tests compare complete stable numerical output."},{"criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"covered","evidence":["src/anomalyzer/io.py:33","tests/test_cli.py:51"],"notes":"The passing CSV regression covers timestamped missing cells, positional data, and varying unmapped entity rejection."},{"criterion":"Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","status":"covered","evidence":["src/anomalyzer/cli.py:253","tests/test_cli.py:51","tests/test_cli.py:71"],"notes":"The passing file/flag/embedded precedence regression and bare-values regression establish the merge order and validated configuration."},{"criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"covered","evidence":["src/anomalyzer/io.py:69","src/anomalyzer/cli.py:424","tests/test_cli.py:41","tests/test_cli.py:108"],"notes":"Output protection and exit regressions passed, including explicit overwrite; successful manual examples have empty stderr and incomplete-usage exits 1."}],"ignore_findings":[],"done":true,"input_fingerprint":"86e701ccc0a1aefd407f11b4547784fef497fa1b49ddf6bb69db0d08f757dfdc","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Run from files or stdin"},"impact":4,"gaps":0,"r":58,"x":903.3333333333334,"y":170},{"id":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","label":"Account for ordinary seaso...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","acceptance":["A supplied season length selects earlier matching seasonal observations for the reference.","Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected."],"guidance":["Expectations support scoring; they are not a standalone forecasting API."],"path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/expectations/protect-reference-from-extremes","analysis/expectations/reset-declared-regimes","analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py"],"automatedChecks":[{"id":"seasonal-reference","description":"Check independent lagged-reference values.","command":"python -m pytest -q tests/test_analysis.py::test_forecast_independent_reference"},{"id":"growth-and-departures","description":"Keep clean linear and compound growth ordinary while identifying injected departures.","command":"python -m pytest -q tests/test_trend.py::test_trend_and_anomalies"},{"id":"frozen-fit","description":"Keep the training fit stable when later observations change.","command":"python -m pytest -q tests/test_trend.py::test_frozen_fit_and_legacy_mode"},{"id":"explicit-trends","description":"Accept supported trend choices and handle short, constant, and signed series.","command":"python -m pytest -q tests/test_trend.py::test_explicit_modes_and_validation tests/test_trend.py::test_short_history_constant_and_signed_linear"}],"manualChecks":["Compare examples/growth_up.json and examples/compound.json with their selected trend and evidence expectations."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","criteria":[{"criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"covered","evidence":["src/anomalyzer/baseline.py:8","src/anomalyzer/baseline.py:31","tests/test_analysis.py:21"],"notes":"Independent lag values and robust matching-phase assertions establish the seasonal reference calculation."},{"criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"covered","evidence":["src/anomalyzer/trend.py:120","src/anomalyzer/trend.py:144","tests/test_trend.py:11"],"notes":"Positive and negative linear/exponential cases select the expected model, match post-training expectations, and produce no ordinary-growth episodes."},{"criterion":"The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","status":"covered","evidence":["src/anomalyzer/recipes.py:162","tests/test_trend.py:36"],"notes":"Trend fitting occurs at the training boundary; the test asserts identical trend diagnostics and earlier evidence after later observations change."},{"criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:20","tests/test_trend.py:49","tests/test_trend.py:57"],"notes":"Tests exercise disabled, linear, exponential, and invalid quadratic modes, including disclosed short-history fallback."}],"ignore_findings":[],"done":true,"input_fingerprint":"8e9a0361b02b7d3f651366e5fa0675cccb3949e44de448d2ae0d43e6710bd408","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Account for growth and seasonality"},"impact":3,"gaps":0,"r":53,"x":170,"y":170},{"id":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","label":"Embed local analysis in my...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","acceptance":["analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","analyze_series preserves supplied settings and event-time context for a single dataset.","analyze_relationships preserves the declared multi-dataset relationship semantics.","Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime."],"guidance":["All interfaces use local analysis; there is no hosted service or model API requirement."],"path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence","interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/__init__.py","src/anomalyzer/agent.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"request-parity","description":"Match canonical request results across Python, CLI, and full-result MCP compatibility calls.","command":"python -m pytest -q tests/test_mcp.py::test_canonical_request_matches_python_cli_and_mcp tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"series-context","description":"Preserve a single-series event-time cutoff and context across interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp tests/test_mcp.py::test_series_wrapper_preserves_omitted_and_explicit_season_settings"},{"id":"relationship-parity","description":"Match declared relationship evidence across the public interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_relationship_semantics_match_python_cli_and_mcp"}],"manualChecks":["Review the Python API examples in README.md and confirm callers can inspect the returned validated models and JSON."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","criteria":[{"criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"covered","evidence":["src/anomalyzer/__init__.py:3","src/anomalyzer/recipes.py:21","tests/test_analysis.py:234"],"notes":"Bare values and separate settings validate before dispatch; canonical schema 1.0 and 1.1 requests use their respective contracts."},{"criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"covered","evidence":["src/anomalyzer/agent.py:11","tests/test_mcp.py:191","tests/test_mcp.py:291"],"notes":"Field-set handling preserves omitted versus explicit settings and context in the single-series wrapper, verified across dictionary and Settings inputs."},{"criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"covered","evidence":["src/anomalyzer/agent.py:27","tests/test_mcp.py:217"],"notes":"Declared relationships validate and use the shared engine without changing units, lineage, or primitive semantics."},{"criterion":"Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime.","status":"covered","evidence":["tests/test_mcp.py:163","tests/test_mcp.py:191","tests/test_mcp.py:217","src/anomalyzer/__main__.py:1"],"notes":"Passing real CLI subprocess, Python, and full-result MCP parity tests compare stable numerical results apart from identity and runtime."}],"ignore_findings":[],"done":true,"input_fingerprint":"214bd10bb3aa5fad0f74938fe2ef09853a593cdb5b93cba49ec851a628d8ab65","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","name":"Python integration","label":"Python integration / Embed analysis in my application / Call the public API"},"impact":3,"gaps":0,"r":53,"x":903.3333333333334,"y":170},{"id":"investigation/cases/group-related-evidence","title":"Investigate related findings together","label":"Investigate related findin...","area":"investigation","scopes":["investigation","investigation/cases"],"scope":"investigation/cases","scopeLabel":"investigation/cases","status":"implemented","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","acceptance":["Evidence for different declared entities remains in separate cases even at the same event time.","Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","A derived relationship adds context without counting as an additional independent source for severity.","Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection."],"guidance":["Warning and critical are deterministic numerical grouping rules, not validated incident impact.","Diagnostic trend, oscillation, mixture, and variation findings remain distinct from adaptive location-shift cases."],"path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","dependencies":["investigation/assessments/describe-early-changes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/mcp/investigate-findings-and-evidence","investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/cases.py","src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py"],"automatedChecks":[{"id":"entity-scope","description":"Separate cases for different entities at the same event time.","command":"python -m pytest -q tests/test_agentic.py::test_cases_are_scoped_by_entity_at_the_same_event_time"},{"id":"source-severity","description":"Escalate numerical severity when two source datasets meet criteria.","command":"python -m pytest -q tests/test_agentic.py::test_two_source_datasets_escalate_case_severity"},{"id":"correlated-context","description":"Keep relationship evidence from double-counting source severity.","command":"python -m pytest -q tests/test_agentic.py::test_relationship_evidence_explains_but_does_not_double_count_severity"},{"id":"first-detection","description":"Group an adaptive location shift before point triggers without using a later pattern peak.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_location_shift_creates_one_early_case_without_point_trigger tests/test_agentic.py::test_location_case_severity_does_not_use_future_pattern_peak"}],"manualChecks":["Inspect a case's contributing targets, explanation, evidence references, and investigation questions; distinguish it from a point-anomaly episode."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","criteria":[{"criterion":"Evidence for different declared entities remains in separate cases even at the same event time.","status":"covered","evidence":["src/anomalyzer/cases.py:42","tests/test_agentic.py:69"],"notes":"Entity participates in grouping and stable identity; the test requires separate cases and distinct IDs at the same time."},{"criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"covered","evidence":["src/anomalyzer/cases.py:72","tests/test_agentic.py:85"],"notes":"Two qualifying source datasets produce critical numerical severity and named source explanations."},{"criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"covered","evidence":["src/anomalyzer/cases.py:72","src/anomalyzer/cases.py:116","tests/test_agentic.py:100"],"notes":"The independent-source count includes dataset criteria only. Derived evidence remains explanatory and can still meet the separately declared magnitude rule."},{"criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"covered","evidence":["src/anomalyzer/early.py:193","src/anomalyzer/early.py:237","src/anomalyzer/orchestrator.py:115","tests/test_agentic.py:122","tests/test_agentic.py:146"],"notes":"Tests require one moderate-shift case without point triggers and restrict assessment references/strength to samples at or before first detection."}],"ignore_findings":[],"done":true,"input_fingerprint":"6830af8a2f064123c12adb062f4793e446b97b3b805fd73a934ea84cd0d5d9cf","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Investigate related evidence"},"impact":3,"gaps":0,"r":53,"x":1086.6666666666667,"y":170},{"id":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","label":"Spot sustained shifts and ...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","acceptance":["Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","Trend, oscillation, and mixture findings retain their specific pattern meanings.","Moving-range findings identify variation changes separately from location changes."],"guidance":["Pattern findings may overlap and are not independent confirmations.","A residual location shift does not establish a change in the full distribution."],"path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["data/quality/analyze-sparse-observations","investigation/cases/group-related-evidence"],"implementationReferences":["src/anomalyzer/evidence.py","src/anomalyzer/recipes.py","tests/test_analysis.py"],"automatedChecks":[{"id":"location-shift","description":"Find moderate increases and decreases without point anomalies.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_patterns_describe_location_shift_without_point_anomaly"},{"id":"accumulating-change","description":"Detect a known CUSUM accumulation at the expected position.","command":"python -m pytest -q tests/test_analysis.py::test_cusum_detects_accumulated_moderate_location_shift"},{"id":"diagnostic-patterns","description":"Distinguish residual trend, oscillation, and mixture rules.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_diagnostics_have_specific_semantics"},{"id":"variation","description":"Describe a moving-range increase as variation rather than location.","command":"python -m pytest -q tests/test_analysis.py::test_moving_range_detects_short_term_variation_increase"}],"manualChecks":["Compare examples/location_shift_up.json with its point episodes and pattern findings, allowing for overlapping residual rules."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","criteria":[{"criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"covered","evidence":["src/anomalyzer/evidence.py:118","src/anomalyzer/evidence.py:189","tests/test_analysis.py:293"],"notes":"Tests assert no point triggers or episodes and exact upward/downward detection positions for Nelson rules and CUSUM."},{"criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"covered","evidence":["src/anomalyzer/evidence.py:211","src/anomalyzer/evidence.py:230","tests/test_analysis.py:350"],"notes":"The known accumulation detects at index 16 and reports peak CUSUM approximately 6, threshold 5, and increase direction."},{"criterion":"Trend, oscillation, and mixture findings retain their specific pattern meanings.","status":"covered","evidence":["src/anomalyzer/evidence.py:70","src/anomalyzer/evidence.py:122","tests/test_analysis.py:340"],"notes":"Separate predicates and descriptions preserve the three meanings and explicitly limit location/distribution claims."},{"criterion":"Moving-range findings identify variation changes separately from location changes.","status":"covered","evidence":["src/anomalyzer/evidence.py:234","tests/test_analysis.py:360"],"notes":"Adjacent standardized-residual differences produce variation_shift findings; the test checks magnitude, detection index, and interpretation."}],"ignore_findings":[],"done":true,"input_fingerprint":"63716eb67db6b73964b79897d8c81f2891df2586ff213c95ea81c5699b8abbea","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Find sustained residual changes"},"impact":2,"gaps":0,"r":48,"x":170,"y":170},{"id":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","label":"Analyze observations in th...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","acceptance":["Calendar datasets provide timestamps and frequency; positional datasets provide neither. Explicit schema-1.1 event-index sources retain timestamps without a calendar frequency.","Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration."],"guidance":["No interpolation or implicit filling, resampling, or unit conversion occurs. Schema-1.1 source policies may explicitly authorize confirmed empty additive volume as zero.","Duplicate null values propagate through aggregation."],"path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["data/quality/exclude-incomplete-periods","relationships/quality/require-compatible-sources"],"implementationReferences":["src/anomalyzer/sparse.py","tests/test_sparse.py","src/anomalyzer/validation.py","src/anomalyzer/contracts.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"coordinate-contract","description":"Reject partial timestamp metadata and positional as_of cutoffs.","command":"python -m pytest -q tests/test_analysis.py::test_timestamp_metadata_is_all_or_nothing"},{"id":"local-time","description":"Normalize valid local times and reject ambiguous or nonexistent times.","command":"python -m pytest -q tests/test_analysis.py::test_local_time"},{"id":"duplicates-cutoff","description":"Require an explicit duplicate policy and retain counts after the cutoff.","command":"python -m pytest -q tests/test_analysis.py::test_duplicate_sort_and_cutoff"},{"id":"cadence-context","description":"Check fixed weekly cadence and preserve cutoff semantics across interfaces.","command":"python -m pytest -q tests/test_analysis.py::test_weekly_cadence tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp"}],"manualChecks":["Inspect a timestamped result's data_quality transformations, UTC window, and resolved context to confirm preparation is explainable."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","criteria":[{"criterion":"Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","status":"covered","evidence":["src/anomalyzer/contracts.py:77","tests/test_analysis.py:418"],"notes":"Partial coordinate metadata and positional as_of are rejected."},{"criterion":"Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","status":"covered","evidence":["src/anomalyzer/validation.py:8","tests/test_analysis.py:500"],"notes":"UTC round trips detect ambiguous/nonexistent local times; valid local time normalizes to the expected UTC hour."},{"criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"covered","evidence":["src/anomalyzer/validation.py:87","src/anomalyzer/validation.py:103","tests/test_analysis.py:478"],"notes":"Explicit mean/sum handling retains duplicate counts and transformation descriptions. Nulls propagate rather than becoming zero."},{"criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"covered","evidence":["src/anomalyzer/validation.py:105","src/anomalyzer/validation.py:111","tests/test_analysis.py:491","tests/test_mcp.py:191"],"notes":"Filtering is inclusive through as_of; cadence compares UTC elapsed timedeltas, including fixed seven-day weeks."}],"ignore_findings":[],"done":true,"input_fingerprint":"791748764cdc0dfe81d987a908179930d858670606e51b64407172295776b261","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Prepare the comparison"},"impact":2,"gaps":0,"r":48,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","label":"Let an agent inspect the e...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","acceptance":["The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","get_evidence exposes the selected finding's numerical support and relationship source lineage.","Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated."],"guidance":["Run results are retained only in the bounded server-session cache; missing or evicted run IDs require reanalysis.","Cases, episodes, and residual patterns may overlap."],"path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":["interfaces/mcp/connect-read-only-local-server","interfaces/skill/guide-anomaly-investigation"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py"],"automatedChecks":[{"id":"case-drilldown","description":"Analyze, page findings, and inspect exact conversion case values and lineage.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case tests/test_mcp.py::test_schema_11_episode_is_paged_and_has_relationship_evidence"},{"id":"unavailable-finding","description":"Expose a quality finding when latest relationship evidence is unavailable.","command":"python -m pytest -q tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"},{"id":"early-candidate","description":"Inspect descriptive candidate evidence without inventing a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case tests/test_mcp.py::test_sparse_insufficient_history_keeps_descriptive_candidate_visible"},{"id":"sample-pagination","description":"Page supporting samples for a residual pattern.","command":"python -m pytest -q tests/test_mcp.py::test_get_evidence_pages_pattern_samples"}],"manualChecks":["Follow analyze, list_findings, and get_evidence for the conversion example and check that the headline is supported by the retrieved values."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","criteria":[{"criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"covered","evidence":["src/anomalyzer/mcp.py:170","src/anomalyzer/mcp_views.py:247","tests/test_mcp.py:54"],"notes":"The conversion manual call returns a relationship headline, preview, counts, readiness, and session-scoped run ID."},{"criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:132","src/anomalyzer/mcp_views.py:260","tests/test_mcp.py:313","tests/test_mcp.py:108"],"notes":"Cases, early candidates, episodes, patterns, and quality findings feed a bounded page; relationship episode regression verifies schema 1.1 pagination."},{"criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:273","src/anomalyzer/mcp_views.py:294","tests/test_mcp.py:54","tests/test_mcp.py:313"],"notes":"Manual drilldown returns observed 8/130 versus expected 0.1 and source indexes orders=14 and qualified_visits=14."},{"criterion":"Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:118","src/anomalyzer/mcp_views.py:169","src/anomalyzer/mcp_views.py:337","tests/test_mcp.py:87","tests/test_mcp.py:137","tests/test_mcp.py:341"],"notes":"Passing tests retain short-history and inapplicable-method candidates plus unavailable latest evidence, while pattern samples are paginated."}],"ignore_findings":[],"done":true,"input_fingerprint":"a09b0003780bcd2f9ccf8e3d8ff58b3147c118760b984757d7027b5810fd9bef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Drill into findings and evidence"},"impact":2,"gaps":0,"r":48,"x":927.3333333333334,"y":288},{"id":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","label":"Spot early changes with a ...","area":"investigation","scopes":["investigation","investigation/assessments"],"scope":"investigation/assessments","scopeLabel":"investigation/assessments","status":"implemented","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","acceptance":["The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","Missing values break the comparison history, and zero previous values do not create relative-change evidence.","Early candidates remain non-triggering and ineligible for action even under a permissive policy."],"guidance":["An early candidate does not establish an anomaly, stable baseline, cause, or incident."],"path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence","investigation/rules/check-declared-numerical-rules"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_mcp.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"prefix-invariant-facts","description":"Describe short history causally without creating cases.","command":"python -m pytest -q tests/test_agentic.py::test_short_history_assessments_are_useful_and_prefix_invariant"},{"id":"comparison-boundaries","description":"Reset comparisons at missing observations and avoid dividing by zero.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"candidate-eligibility","description":"Keep candidates ineligible for action and notifications under permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"},{"id":"agent-visibility","description":"Expose a short-history candidate and its claim limits without a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case"}],"manualChecks":["Inspect a candidate's comparison population, maturity, established claim, and not_established text before describing it."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","criteria":[{"criterion":"The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","status":"covered","evidence":["src/anomalyzer/early.py:73","src/anomalyzer/early.py:96","tests/test_agentic.py:35"],"notes":"Assertions establish observation/change/change/departure_candidate for [10,12,11,20]."},{"criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"covered","evidence":["src/anomalyzer/early.py:50","src/anomalyzer/early.py:141","tests/test_agentic.py:35"],"notes":"Each assessment precedes the prior-history update; tests compare complete serialized assessments for every prefix."},{"criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"covered","evidence":["src/anomalyzer/early.py:53","src/anomalyzer/early.py:58","tests/test_capability_behaviors.py:87"],"notes":"The zero comparison has no relative change; after a missing value, comparison samples reset to zero."},{"criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"covered","evidence":["src/anomalyzer/early.py:105","src/anomalyzer/cases.py:35","tests/test_capability_behaviors.py:134","tests/test_mcp.py:341"],"notes":"Candidates have criterion_met=false and remain action/notification-ineligible after policy replay, without creating cases."}],"ignore_findings":[],"done":true,"input_fingerprint":"debdeeeaea559950f53461a4298cb9c2e3a1681f83a43e8a981ed9b51968a4d0","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect early evidence"},"impact":2,"gaps":0,"r":48,"x":1110.6666666666667,"y":288},{"id":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","label":"Compare the same entity an...","area":"relationships","scopes":["relationships","relationships/quality"],"scope":"relationships/quality","scopeLabel":"relationships/quality","status":"implemented","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","acceptance":["Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","Difference relationships require identical declared source units and matching output units.","An invalid relationship is reported as failed while valid sibling dataset results remain available."],"guidance":["A failed or unavailable relationship does not establish that its metric is normal.","No implicit timestamp matching, interpolation, unit conversion, or zero-filling occurs."],"path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"implementationReferences":["src/anomalyzer/alignment.py","src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"exact-alignment","description":"Reject mismatched timestamps, cadence, entity, duplicates, and positional relationship sources.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relationship_requires_exact_source_alignment"},{"id":"compatible-units","description":"Report a difference unit mismatch without suppressing sibling results.","command":"python -m pytest -q tests/test_agentic.py::test_alignment_unit_and_partial_failure_are_specific"},{"id":"unambiguous-declaration","description":"Reject fields that do not belong to the selected relationship kind.","command":"python -m pytest -q tests/test_agentic.py::test_contract_rejects_ambiguous_relationship_fields"},{"id":"healthy-siblings","description":"Retain a healthy dataset's evidence when another dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Read both target-level statuses and errors before interpreting a partial run; check source dataset errors when a relationship source failed."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","criteria":[{"criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"covered","evidence":["src/anomalyzer/alignment.py:42","tests/test_capability_behaviors.py:30"],"notes":"Parameterized assertions require specific errors for timestamp, cadence, and entity mismatches."},{"criterion":"Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","status":"covered","evidence":["src/anomalyzer/alignment.py:23","src/anomalyzer/alignment.py:44","tests/test_capability_behaviors.py:30"],"notes":"Positional sources fail exact alignment; even an explicit duplicate-mean setting is rejected for relationship source lineage."},{"criterion":"Difference relationships require identical declared source units and matching output units.","status":"covered","evidence":["src/anomalyzer/relationships.py:41","tests/test_agentic.py:242"],"notes":"Source and output unit checks are explicit. Tests cover mismatched source units and successful matching units."},{"criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:171","tests/test_capability_behaviors.py:30","tests/test_capability_behaviors.py:159"],"notes":"Relationship errors remain target-specific, healthy dataset results survive, and the overall mixed run is partial."}],"ignore_findings":[],"done":true,"input_fingerprint":"06321ab5806a8276cc3954df5d6f4dd7f059f67600038854569787ff51b420d6","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Align comparable sources"},"impact":2,"gaps":0,"r":48,"x":1294,"y":288},{"id":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","label":"Compare against a new base...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","acceptance":["Integer positions and exact observed dates or timestamps can identify reset boundaries.","Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","Repeated, out-of-range, or unmatched reset points are rejected.","CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries."],"guidance":["Reset points are explicit caller decisions. Known-event notes and review-state fields do not rebuild regimes.","Multi-resolution analysis does not support reset_points."],"path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["data/quality/analyze-sparse-observations"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py"],"automatedChecks":[{"id":"independent-regimes","description":"Retrain separate regimes without cross-boundary evidence or patterns.","command":"python -m pytest -q tests/test_analysis.py::test_manual_resets_retrain_without_crossing_boundaries"},{"id":"invalid-boundaries","description":"Reject invalid or duplicate resolved reset positions.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_manual_resets"},{"id":"cli-boundaries","description":"Resolve position and date reset flags through the CLI.","command":"python -m pytest -q tests/test_cli.py::test_manual_reset_cli"}],"manualChecks":["Inspect the resolved reset positions and segment diagnostics for examples/steps_reset.json."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","criteria":[{"criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"covered","evidence":["src/anomalyzer/validation.py:129","tests/test_analysis.py:381"],"notes":"Positions and observed dates are exercised end to end; timestamp normalization and exact matching are implemented directly."},{"criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"covered","evidence":["src/anomalyzer/recipes.py:119","src/anomalyzer/recipes.py:143","tests/test_analysis.py:381"],"notes":"Each segment initializes separate state and passes its own start to forecasts. Tests verify three calibrated regimes and no cross-boundary evidence or findings."},{"criterion":"Repeated, out-of-range, or unmatched reset points are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/validation.py:149","tests/test_analysis.py:413"],"notes":"Validation rejects literal duplicates and different declarations resolving to the same position, invalid positions, and unmatched dates/timestamps."},{"criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"covered","evidence":["tests/test_cli.py:82","src/anomalyzer/recipes.py:53"],"notes":"CLI assertions establish positions [50, 100] and removal of the unmarked-regime episodes."}],"ignore_findings":[],"done":true,"input_fingerprint":"067063f972c5b70d990d6cb0125892731a97a7f482698d7b331a1a30229ac223","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Restart after a reviewed change"},"impact":1,"gaps":0,"r":43,"x":194,"y":288},{"id":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","label":"Assess sparse observations...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Schema-1.1 callers declare coverage and missing-data semantics per source, keeping unknown collection distinct from confirmed empty additive volume.","intent":"Retain useful rule and supported statistical evidence for sparse histories while preserving time, provenance, incomplete periods and evidence limits.","acceptance":["Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals."],"guidance":["Default missingness is unknown. A modeled expectation is never a measurement.","Caller business severity, incident state and notifications remain outside the engine."],"path":".capabilities/data/quality/analyze-sparse-observations.capability.yaml","dependencies":["data/inputs/accept-analysis-requests","analysis/expectations/reset-declared-regimes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["relationships/resolution/compare-finalized-subperiods"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/sparse.py","src/anomalyzer/sparse_views.py","src/anomalyzer/orchestrator.py","src/anomalyzer/relationships.py","src/anomalyzer/early.py","tests/test_sparse.py","docs/sparse-data.md","examples/sparse_examples.py","examples/render_readme_progression.py"],"automatedChecks":[{"id":"source-semantics","description":"Preserve unknown periods and require explicit empty-volume authorization.","command":"python -m pytest -q tests/test_sparse.py -k \"zero or omitted or grid or duplicate or expansion\""},{"id":"causal-boundaries","description":"Retain real phase, event times, resets and as-of revision eligibility.","command":"python -m pytest -q tests/test_sparse.py -k \"phase or gap or reset or event or revision or arrival\""},{"id":"public-interfaces","description":"Preserve useful rules, relationships, views, cases and API parity.","command":"python -m pytest -q tests/test_sparse.py -k \"rule or relationship or daily or parity\""},{"id":"progressive-examples","description":"Reproduce the full numbered sequence and sparse worked examples.","command":"python examples/run_demo.py"}],"manualChecks":["Inspect progression figures 14–22 for true gaps, distinguished policy zeros, original event time, unavailable checks and caller-owned business severity."],"verificationGaps":[],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for data/quality/analyze-sparse-observations.","criteria":[{"criterion":"Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","status":"partial","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:71","src/anomalyzer/contracts.py:85","src/anomalyzer/contracts.py:86","src/anomalyzer/sparse.py:3","src/anomalyzer/sparse.py:34","src/anomalyzer/sparse.py:42","src/anomalyzer/sparse.py:48","src/anomalyzer/sparse.py:54","src/anomalyzer/sparse_views.py:74","src/anomalyzer/sparse_views.py:81","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:97","src/anomalyzer/sparse_views.py:119","src/anomalyzer/orchestrator.py:63","src/anomalyzer/orchestrator.py:127","src/anomalyzer/orchestrator.py:139","src/anomalyzer/orchestrator.py:191","src/anomalyzer/relationships.py:36","src/anomalyzer/relationships.py:58","src/anomalyzer/relationships.py:98","src/anomalyzer/relationships.py:131","src/anomalyzer/relationships.py:141","src/anomalyzer/early.py:58","src/anomalyzer/early.py:137","tests/test_sparse.py:44","tests/test_sparse.py:45","tests/test_sparse.py:50","tests/test_sparse.py:51","tests/test_sparse.py:57","docs/sparse-data.md:4","docs/sparse-data.md:29","docs/sparse-data.md:37","docs/sparse-data.md:39","docs/sparse-data.md:40","examples/sparse_examples.py:22","examples/sparse_examples.py:23","examples/sparse_examples.py:25","examples/sparse_examples.py:26","examples/sparse_examples.py:28","examples/render_readme_progression.py:135","examples/render_readme_progression.py:139","examples/render_readme_progression.py:140","examples/render_readme_progression.py:166","examples/render_readme_progression.py:206"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","status":"partial","evidence":["src/anomalyzer/contracts.py:27","src/anomalyzer/contracts.py:49","src/anomalyzer/contracts.py:51","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:56","src/anomalyzer/sparse.py:1","src/anomalyzer/sparse.py:23","src/anomalyzer/sparse.py:181","src/anomalyzer/sparse.py:187","src/anomalyzer/sparse.py:189","src/anomalyzer/sparse_views.py:15","src/anomalyzer/sparse_views.py:98","src/anomalyzer/sparse_views.py:107","src/anomalyzer/sparse_views.py:108","src/anomalyzer/sparse_views.py:112","src/anomalyzer/orchestrator.py:22","src/anomalyzer/orchestrator.py:23","src/anomalyzer/orchestrator.py:38","src/anomalyzer/orchestrator.py:129","src/anomalyzer/orchestrator.py:141","src/anomalyzer/relationships.py:185","src/anomalyzer/relationships.py:189","src/anomalyzer/relationships.py:191","src/anomalyzer/early.py:46","src/anomalyzer/early.py:48","src/anomalyzer/early.py:51","src/anomalyzer/early.py:59","src/anomalyzer/early.py:61","tests/test_sparse.py:20","tests/test_sparse.py:108","tests/test_sparse.py:111","tests/test_sparse.py:113","tests/test_sparse.py:119","docs/sparse-data.md:37","docs/sparse-data.md:41","docs/sparse-data.md:42","docs/sparse-data.md:106","docs/sparse-data.md:108","examples/sparse_examples.py:16","examples/sparse_examples.py:33","examples/sparse_examples.py:50","examples/sparse_examples.py:51","examples/sparse_examples.py:91","examples/render_readme_progression.py:241","examples/render_readme_progression.py:312","examples/render_readme_progression.py:317","examples/render_readme_progression.py:346","examples/render_readme_progression.py:349"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","status":"partial","evidence":["src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:81","src/anomalyzer/contracts.py:242","src/anomalyzer/contracts.py:245","src/anomalyzer/contracts.py:258","src/anomalyzer/sparse.py:4","src/anomalyzer/sparse.py:49","src/anomalyzer/sparse.py:57","src/anomalyzer/sparse.py:63","src/anomalyzer/sparse.py:64","src/anomalyzer/sparse_views.py:75","src/anomalyzer/sparse_views.py:94","src/anomalyzer/sparse_views.py:127","src/anomalyzer/sparse_views.py:155","src/anomalyzer/sparse_views.py:157","src/anomalyzer/orchestrator.py:37","src/anomalyzer/orchestrator.py:38","src/anomalyzer/orchestrator.py:166","src/anomalyzer/orchestrator.py:178","src/anomalyzer/orchestrator.py:180","src/anomalyzer/relationships.py:40","src/anomalyzer/relationships.py:57","src/anomalyzer/relationships.py:58","src/anomalyzer/relationships.py:96","src/anomalyzer/relationships.py:97","src/anomalyzer/early.py:140","src/anomalyzer/early.py:147","tests/test_sparse.py:59","tests/test_sparse.py:67","tests/test_sparse.py:75","tests/test_sparse.py:99","tests/test_sparse.py:128","docs/sparse-data.md:26","docs/sparse-data.md:96","docs/sparse-data.md:101","docs/sparse-data.md:132","docs/sparse-data.md:142","examples/sparse_examples.py:26","examples/sparse_examples.py:36","examples/sparse_examples.py:38","examples/sparse_examples.py:57","examples/sparse_examples.py:81","examples/render_readme_progression.py:654","examples/render_readme_progression.py:697","examples/render_readme_progression.py:884","examples/render_readme_progression.py:896","examples/render_readme_progression.py:917"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals.","status":"partial","evidence":["src/anomalyzer/contracts.py:13","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:297","src/anomalyzer/contracts.py:298","src/anomalyzer/contracts.py:437","src/anomalyzer/sparse.py:4","src/anomalyzer/sparse.py:34","src/anomalyzer/sparse.py:64","src/anomalyzer/sparse.py:66","src/anomalyzer/sparse.py:83","src/anomalyzer/sparse_views.py:1","src/anomalyzer/sparse_views.py:28","src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:94","src/anomalyzer/orchestrator.py:106","src/anomalyzer/orchestrator.py:107","src/anomalyzer/orchestrator.py:133","src/anomalyzer/orchestrator.py:166","src/anomalyzer/orchestrator.py:180","src/anomalyzer/relationships.py:160","src/anomalyzer/relationships.py:162","src/anomalyzer/relationships.py:166","tests/test_sparse.py:59","tests/test_sparse.py:75","tests/test_sparse.py:99","tests/test_sparse.py:123","tests/test_sparse.py:180","docs/sparse-data.md:26","docs/sparse-data.md:41","docs/sparse-data.md:69","docs/sparse-data.md:78","docs/sparse-data.md:96","examples/sparse_examples.py:59","examples/sparse_examples.py:95","examples/render_readme_progression.py:686","examples/render_readme_progression.py:699","examples/render_readme_progression.py:710","examples/render_readme_progression.py:767","examples/render_readme_progression.py:884"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","partial: Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","partial: Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","partial: As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals."],"reviewStale":false,"health":"gap","healthLabel":"Review findings","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Assess sparse source data","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Assess sparse source data"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","label":"Avoid mistaking unfinished...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Explicitly incomplete observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","acceptance":["Schema-1.0 incomplete observations form a trailing suffix; schema-1.1 source policies preserve internal and trailing incomplete positions as unavailable.","The observed values and timestamps of unfinished periods remain available in data quality output.","Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","A relationship with an unavailable latest source reports that unavailability rather than normal behavior."],"guidance":["Completeness is declared by the caller; the engine does not infer collection completeness."],"path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/sparse.py","src/anomalyzer/sparse_views.py","tests/test_sparse.py","src/anomalyzer/validation.py","src/anomalyzer/alignment.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"unfinished-values","description":"Retain an unfinished value without scoring it.","command":"python -m pytest -q tests/test_analysis.py::test_explicit_incomplete_period_is_reported_and_not_scored"},{"id":"trailing-only","description":"Reject incomplete observations followed by completed observations.","command":"python -m pytest -q tests/test_analysis.py::test_incomplete_periods_must_be_trailing"},{"id":"source-and-agent-views","description":"Exclude a partial source sample from both analysis views and expose unavailable relationship evidence.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"}],"manualChecks":["Inspect examples/agentic/incomplete-usage.json and confirm the latest headline explains missing completeness and does not imply normality."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","criteria":[{"criterion":"Incomplete observations form a trailing suffix after chronological preparation.","status":"covered","evidence":["src/anomalyzer/validation.py:44","tests/test_analysis.py:162"],"notes":"Suffix validation runs after timestamp preparation and rejects a completed observation following an incomplete one."},{"criterion":"The observed values and timestamps of unfinished periods remain available in data quality output.","status":"covered","evidence":["src/anomalyzer/validation.py:50","src/anomalyzer/validation.py:114","tests/test_analysis.py:150"],"notes":"Quality records retain original source index, normalized timestamp, observed value, and incomplete status."},{"criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"covered","evidence":["src/anomalyzer/validation.py:53","src/anomalyzer/recipes.py:564","tests/test_analysis.py:150","tests/test_analysis.py:200"],"notes":"Only finalized prepared values enter either scoring path; assertions verify the last scored sample and both multi-resolution counts."},{"criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"covered","evidence":["src/anomalyzer/alignment.py:35","src/anomalyzer/orchestrator.py:150","tests/test_mcp.py:87"],"notes":"Incomplete aligned sources become unavailable values, and latest relationship state/status plus MCP headline explain the lack of evidence."}],"ignore_findings":[],"done":true,"input_fingerprint":"d82ee816fe89a25232e2589587d31bcadfa81be4188a0e1531f4fbfdf0ebcb5e","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Keep unfinished data visible"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","label":"Give an agent read-only ac...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","acceptance":["Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","Analysis tool calls return structured content together with compact text.","Invalid tool input and unknown session run IDs return structured errors rather than fabricated results."],"guidance":["Compatibility tool names remain callable but are not advertised.","Legacy clients receive isError tool results; modern clients receive JSON-RPC invalid-input errors.","The server provides analysis and evidence inspection; the caller owns monitoring and operational workflows."],"path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","dependencies":["interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":["interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py","pyproject.toml"],"automatedChecks":[{"id":"tool-discovery","description":"Advertise the three read-only tools and their schema contracts.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_lists_read_only_tools"},{"id":"connection-protocols","description":"Support legacy stdio initialization and modern stateless discovery.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_stdio_legacy_handshake_and_tool_listing tests/test_mcp.py::test_mcp_modern_discovery_and_stateless_tool_listing"},{"id":"structured-content","description":"Return a compact analysis response with structured content and run ID.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_tool_call_returns_structured_content"},{"id":"invalid-calls","description":"Reject malformed input and unknown session run IDs in both protocols.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_invalid_tool_input_is_a_tool_error tests/test_mcp.py::test_paged_tool_inputs_enforce_schema_in_both_protocols"}],"manualChecks":["Review discovery's interpretation guidance and confirm the client can start anomalyzer-mcp in the installed Python environment."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","criteria":[{"criterion":"Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","status":"covered","evidence":["src/anomalyzer/mcp.py:63","tests/test_mcp.py:27"],"notes":"Discovery exposes three read-only tools with the same strict argument contracts used at dispatch and advertised output schemas."},{"criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"covered","evidence":["src/anomalyzer/mcp.py:228","tests/test_mcp.py:248","tests/test_mcp.py:268"],"notes":"Legacy initialization and modern stateless discovery tests establish protocol metadata and evidence interpretation guidance."},{"criterion":"Analysis tool calls return structured content together with compact text.","status":"covered","evidence":["src/anomalyzer/mcp.py:162","tests/test_mcp.py:39"],"notes":"The manual conversion call returns a compact headline and structured run ID, counts, readiness, and preview."},{"criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"covered","evidence":["src/anomalyzer/mcp.py:156","src/anomalyzer/mcp.py:257","tests/test_mcp.py:235","tests/test_mcp.py:367"],"notes":"Tests reject invalid cursor/limit types, unknown fields, numeric run IDs, and missing session runs in both protocols without fabricated results."}],"ignore_findings":[],"done":true,"input_fingerprint":"7bd92d238f4f9545372c6c35e78aefea60c585bea0a4a8bb99096ab6cef27f96","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Connect a local server"},"impact":1,"gaps":0,"r":43,"x":903.3333333333334,"y":406},{"id":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","label":"Find values that break my ...","area":"investigation","scopes":["investigation","investigation/rules"],"scope":"investigation/rules","scopeLabel":"investigation/rules","status":"implemented","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","acceptance":["Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","Absolute and relative changes must exceed their declared maximum to violate a rule.","Relative-change checks require a nonzero previous value and do not bridge missing observations.","Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them."],"guidance":["An exact rule crossing does not establish cause, an incident, or business impact."],"path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","dependencies":["investigation/assessments/describe-early-changes"],"suggestedDependencies":[],"dependents":["investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"rule-boundaries","description":"Check inclusive range boundaries and strict absolute/relative change thresholds.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_explicit_rules_respect_boundaries"},{"id":"relative-availability","description":"Avoid relative violations across zero or missing values.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"explicit-not-actionable","description":"Create evidence and a case for an exact rule crossing without enabling action by default.","command":"python -m pytest -q tests/test_agentic.py::test_explicit_rule_is_exact_but_not_actionable_without_policy"},{"id":"readable-rule","description":"Identify the exact unit-cost range rule in CLI output.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the unit-cost example's crossed rule and confirm its text distinguishes the declared bound from statistical detection."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","criteria":[{"criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"covered","evidence":["src/anomalyzer/early.py:22","src/anomalyzer/contracts.py:209","tests/test_capability_behaviors.py:74"],"notes":"Boundary values 10 and 20 are accepted; 9 and 21 violate the declared range."},{"criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"covered","evidence":["src/anomalyzer/early.py:25","tests/test_capability_behaviors.py:74"],"notes":"Parameterized tests distinguish equality to the threshold from strict exceedance for both change rules."},{"criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"covered","evidence":["src/anomalyzer/early.py:29","src/anomalyzer/early.py:53","tests/test_capability_behaviors.py:87"],"notes":"Zero/missing predecessors suppress relative comparisons, and missing values reset comparison history."},{"criterion":"Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them.","status":"covered","evidence":["src/anomalyzer/early.py:34","src/anomalyzer/early.py:121","tests/test_agentic.py:49","tests/test_cli.py:134"],"notes":"Violation records retain rule ID/kind/bounds; default eligibility is false and replay requires explicit permission."}],"ignore_findings":[],"done":true,"input_fingerprint":"60d24e84faee9c1e2ae8f9337d17334b37c89305093404da2e84c51526111d0b","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Apply my stated limits"},"impact":1,"gaps":0,"r":43,"x":1086.6666666666667,"y":406},{"id":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","label":"Find changes in conversion...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","acceptance":["Ratios retain the computed value, formula, and original source indexes for each sample.","Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","Derived series can be analyzed with the frozen or adaptive single-resolution recipe."],"guidance":["Relationships are caller-supplied declarations, not discovered or causal relationships.","A difference requires identical declared units; a normalized residual requires a positive supplied scale.","A joint condition is a declared 0/1 comparison, not statistical proof."],"path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","dependencies":["relationships/quality/require-compatible-sources","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["relationships/resolution/compare-finalized-subperiods"],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_examples.py"],"automatedChecks":[{"id":"ratio-lineage","description":"Explain a changing rate with the exact source indexes and formula.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"other-comparisons","description":"Check declared differences, normalized residuals, and combined conditions.","command":"python -m pytest -q tests/test_agentic.py::test_remaining_relationship_primitives"},{"id":"unavailable-values","description":"Keep zero-denominator and missing-source observations unavailable.","command":"python -m pytest -q tests/test_agentic.py::test_missing_and_zero_denominator_are_unavailable_not_zero"},{"id":"adaptive-relationship","description":"Use the adaptive recipe for a declared relationship.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_recipe_runs_for_declared_relationships"}],"manualChecks":["Inspect examples/agentic/conversion.json and distinguish its relationship departure from the individual source metrics."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","criteria":[{"criterion":"Ratios retain the computed value, formula, and original source indexes for each sample.","status":"covered","evidence":["src/anomalyzer/relationships.py:84","src/anomalyzer/relationships.py:152","tests/test_agentic.py:211"],"notes":"Assertions check exact ratio values, formula, and both original source indexes."},{"criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"covered","evidence":["src/anomalyzer/relationships.py:97","src/anomalyzer/relationships.py:134","tests/test_agentic.py:285"],"notes":"Parameterized assertions independently specify difference, normalized residual, all-condition, and any-condition outputs."},{"criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"covered","evidence":["src/anomalyzer/relationships.py:89","src/anomalyzer/relationships.py:142","tests/test_agentic.py:226"],"notes":"Lineage retains null values with missing_source_value/zero_denominator reasons. Explicit zero_denominator=error intentionally fails instead, as documented."},{"criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:87","src/anomalyzer/orchestrator.py:135","tests/test_agentic.py:171","tests/test_examples.py:17"],"notes":"Both supported recipes dispatch derived datasets through the shared single-resolution engine."}],"ignore_findings":[],"done":true,"input_fingerprint":"7fa8f7ffdc0d852ad961b3e1753761d1bd3b00b87b8c001a957ec19e26d068b9","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Compare ratios and metric gaps"},"impact":1,"gaps":0,"r":43,"x":1270,"y":406},{"id":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","label":"Find unusual spikes and dr...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","acceptance":["A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","Separated point departures remain separate episodes, and positional episode intervals use integer coordinates."],"guidance":["Detection is a numerical departure from the selected expectation, not a claim about cause or business impact."],"path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/evidence.py","tests/test_analysis.py"],"automatedChecks":[{"id":"point-evidence","description":"Detect an injected spike and retain deterministic earlier numerical evidence.","command":"python -m pytest -q tests/test_analysis.py::test_spike_and_no_lookahead"},{"id":"adjacent-run","description":"Describe adjacent point triggers and their contributing positions.","command":"python -m pytest -q tests/test_analysis.py::test_anomaly_stream_detects_consecutive_run"},{"id":"isolated-events","description":"Keep separated anomalies from becoming a consecutive-run pattern.","command":"python -m pytest -q tests/test_analysis.py::test_isolated_anomalies_do_not_form_second_order_pattern"},{"id":"positional-episodes","description":"Use integer intervals for position-only episode evidence.","command":"python -m pytest -q tests/test_analysis.py::test_positional_episode_uses_integer_interval"}],"manualChecks":["Inspect observations and triggering_samples for examples/spike.json and distinguish episode grouping from pattern findings."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","criteria":[{"criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"covered","evidence":["src/anomalyzer/recipes.py:344","src/anomalyzer/recipes.py:388","tests/test_analysis.py:441"],"notes":"The injected spike retains its observed value, expectation, standardized residual, and point trigger."},{"criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"covered","evidence":["src/anomalyzer/baseline.py:31","src/anomalyzer/recipes.py:174","tests/test_analysis.py:441","tests/test_trend.py:36"],"notes":"Assertions establish repeatability and unchanged earlier evidence under later evaluation changes with fixed seasonal settings. Automatic positional season discovery has the separately documented streaming limitation."},{"criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"covered","evidence":["src/anomalyzer/evidence.py:95","tests/test_analysis.py:274"],"notes":"The test checks one three-sample run, exact contributing positions, direction, and first detection."},{"criterion":"Separated point departures remain separate episodes, and positional episode intervals use integer coordinates.","status":"covered","evidence":["src/anomalyzer/evidence.py:12","src/anomalyzer/evidence.py:23","tests/test_analysis.py:263","tests/test_analysis.py:369"],"notes":"Separated spikes produce two episodes without a consecutive-run pattern; positional evidence produces interval [9, 9]."}],"ignore_findings":[],"done":true,"input_fingerprint":"062833cc1bd19d5734b0f3c63d1e81666ffb568ead3399e359d62db2c0c0677d","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Find spikes and drops"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","label":"Know when the expected bas...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","acceptance":["Repeating seasonal structure with stable calibration can yield supported readiness.","Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","Insufficient calibration support yields not_assessed, including when the run otherwise completes.","Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment."],"guidance":["Readiness does not prove real-world effectiveness and is not an automatic action-policy gate."],"path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/readiness.py","src/anomalyzer/recipes.py","tests/test_readiness.py"],"automatedChecks":[{"id":"supported-reference","description":"Support repeated seasonality and preserve readiness under later changes.","command":"python -m pytest -q tests/test_readiness.py::test_repeating_season_and_trend_support_reference"},{"id":"weak-season","description":"Warn on unstable seasonality even when no findings occur.","command":"python -m pytest -q tests/test_readiness.py::test_unstable_declared_season_warns_without_calling_series_inapplicable tests/test_readiness.py::test_documented_random_example_has_no_findings_but_warns_on_reference"},{"id":"calibration-drift","description":"Warn when residual calibration drifts despite a repeated training season.","command":"python -m pytest -q tests/test_readiness.py::test_calibration_drift_warns_even_when_training_season_repeats"},{"id":"not-assessed","description":"Distinguish uncalibrated history and too few calibration samples.","command":"python -m pytest -q tests/test_readiness.py::test_readiness_is_not_assessed_without_calibrated_evaluation tests/test_readiness.py::test_short_calibration_is_not_assessed_even_when_analysis_completes"}],"manualChecks":["Read readiness reasons alongside findings; a lack of findings alone does not establish a trustworthy reference."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","criteria":[{"criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"covered","evidence":["src/anomalyzer/readiness.py:128","tests/test_readiness.py:8"],"notes":"Assertions establish repeated seasonality, strong calibration improvement, and supported readiness."},{"criterion":"Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","status":"covered","evidence":["src/anomalyzer/readiness.py:113","src/anomalyzer/readiness.py:122","tests/test_readiness.py:28","tests/test_readiness.py:41"],"notes":"Completed analyses retain caution reasons for weak seasonal repetition or calibration-center drift."},{"criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"covered","evidence":["src/anomalyzer/readiness.py:77","tests/test_readiness.py:56","tests/test_readiness.py:63"],"notes":"Both uncalibrated history and a completed run with three calibration samples return not_assessed."},{"criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"covered","evidence":["src/anomalyzer/recipes.py:327","tests/test_readiness.py:8"],"notes":"Readiness reads training/calibration inputs once; tests compare identical segment and aggregate assessments after a later shift."}],"ignore_findings":[],"done":true,"input_fingerprint":"d01367627e058ea39c78ee0e3616061a2e0725d06e8acd0b84b46a9655431e57","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Assess baseline quality"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","label":"Follow evolving growth whi...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","acceptance":["Each rolling expectation uses only observations before the sample being scored.","Persistent growth changes can be followed more closely than by the frozen default recipe.","An isolated extreme value is held out of future model references rather than creating a seasonal echo.","Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected."],"guidance":["A persistent third extreme miss in one direction may be admitted to the rolling reference; this is not a manual regime reset."],"path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"causal-window","description":"Keep a known rolling expectation invariant when future values are appended.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_expectation_is_one_step_and_prefix_invariant"},{"id":"changing-growth","description":"Follow a persistent growth change and report model admissions.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_tracks_a_persistent_growth_change"},{"id":"isolated-extreme","description":"Exclude an isolated extreme from later reference estimates.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_contains_an_isolated_extreme"},{"id":"bounded-settings","description":"Reject conflicting or unbounded adaptive settings.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_settings_reject_conflicting_or_unbounded_trend"}],"manualChecks":["Inspect the adaptive recipe's rolling-window parameters and frozen calibration before interpreting its final expectation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","criteria":[{"criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"covered","evidence":["src/anomalyzer/baseline.py:61","tests/test_analysis.py:48"],"notes":"All rolling-reference positions precede the scored index. The independent expectation remains unchanged after future values are appended."},{"criterion":"Persistent growth changes can be followed more closely than by the frozen default recipe.","status":"covered","evidence":["src/anomalyzer/recipes.py:145","src/anomalyzer/recipes.py:370","tests/test_analysis.py:58"],"notes":"The test requires adaptive held-out MAE below half the frozen recipe's MAE and a close final expectation."},{"criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","tests/test_analysis.py:81"],"notes":"The isolated spike is excluded from model references; its next seasonal expectation closely matches the control. Persistent third misses are admitted as guidance explicitly permits."},{"criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:41","tests/test_analysis.py:104"],"notes":"Both incompatible explicit trend selection and excessive slope lookback raise specific validation errors."}],"ignore_findings":[],"done":true,"input_fingerprint":"984422f1fd3ab3d22cf7b4ba9d0bd06629fc228a8e53d7ad44439cd8d35ec0db","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Follow evolving growth"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","label":"Account for daily and week...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"planned","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","acceptance":["Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","Insufficient support for a requested cycle produces an explicit limitation or declared fallback rather than a silently fitted component.","Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle."],"guidance":["Begin with a bounded daily-and-weekly use case; overlapping cycles share information and cannot be treated as independent evidence.","Proposed automated checks cover combined ordinary cycles, future isolation, unsupported cycle history, and extreme-value influence across cycles.","Manual review should inspect a weekday morning spike and verify the baseline explains ordinary calendar variation without explaining away the spike."],"path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Represent more than one repeating cycle"},"impact":0,"gaps":3,"r":47,"x":170,"y":406},{"id":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","label":"Limit how extreme values d...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","acceptance":["Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","Training and calibration diagnostics identify excluded extreme observations.","An extreme evaluated sample remains evidence while its expected value is used in future references.","Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior."],"guidance":["Protection keeps the observed evidence; it does not delete observations or certify that a departure was an incident."],"path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/robust.py","src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py","tests/test_cli.py"],"automatedChecks":[{"id":"robust-reference","description":"Limit an extreme weekly reference and use earlier references at lag one.","command":"python -m pytest -q tests/test_analysis.py::test_robust_weekly_reference_limits_one_high_week tests/test_analysis.py::test_robust_lag_one_uses_prior_seasonal_values"},{"id":"training-calibration","description":"Record training and calibration exclusions and retain a useful growth fit.","command":"python -m pytest -q tests/test_trend.py::test_robust_training_fit_and_calibration_exclusions"},{"id":"include-choice","description":"Compare robust and include modes through the CLI and show their differing triggers.","command":"python -m pytest -q tests/test_cli.py::test_outlier_handling_cli"}],"manualChecks":["Inspect excluded_from_model, reference_action, and outlier-handling diagnostics alongside the original observed value."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","criteria":[{"criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"covered","evidence":["src/anomalyzer/baseline.py:39","tests/test_analysis.py:25"],"notes":"Assertions check a independently computed weighted reference and invariance when an already-trimmed high value becomes more extreme."},{"criterion":"Training and calibration diagnostics identify excluded extreme observations.","status":"covered","evidence":["src/anomalyzer/trend.py:178","src/anomalyzer/recipes.py:287","tests/test_trend.py:65"],"notes":"Tests identify exact training/calibration exclusions and compare their influence on trend selection and residual scale."},{"criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","src/anomalyzer/recipes.py:392","tests/test_analysis.py:263","tests/test_trend.py:11"],"notes":"The observed departure and trigger survive; only reference_values receives the prior expectation, preventing the next seasonal echo."},{"criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"covered","evidence":["src/anomalyzer/recipes.py:151","src/anomalyzer/recipes.py:475","tests/test_cli.py:95"],"notes":"CLI assertions distinguish robust triggers {65} from include-mode triggers {65, 72}."}],"ignore_findings":[],"done":true,"input_fingerprint":"183b8696fdc9786878b3e54c4a9e5f68180b47e157528bef953274a3487d49a7","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Protect later comparisons"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","label":"Choose the repeating cycle...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","acceptance":["An explicit season length is honored and reported as explicit.","A repeating positional series can infer a seasonal lag and report the selection window.","A positional series without a sufficiently stable candidate falls back to lag one and records why.","The reserved calibration and evaluation tail does not determine positional season selection."],"guidance":["Candidate discovery uses the input prefix before the reserved tail and confirms the candidate in an early window. It is not fully prefix-invariant streaming discovery.","Supply season_length explicitly when stable streaming season selection is required."],"path":".capabilities/analysis/expectations/select-season-length.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/seasonality.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"declared-season","description":"Honor a separate explicit season setting.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"inference-fallback","description":"Infer a repeating lag or record lag-one fallback when no season is supported.","command":"python -m pytest -q tests/test_analysis.py::test_positional_season_inference_and_fallback"},{"id":"reserved-tail","description":"Keep selection unchanged when only the reserved tail changes.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_season_inference_excludes_reserved_tail"},{"id":"cli-inference","description":"Expose the inferred lag and its source for bare JSON values in the CLI.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"}],"manualChecks":["Review selection_window and ranked candidates for a positional series before treating its inferred season as an ongoing monitoring assumption."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","criteria":[{"criterion":"An explicit season length is honored and reported as explicit.","status":"covered","evidence":["src/anomalyzer/recipes.py:55","tests/test_analysis.py:234"],"notes":"Explicit field presence bypasses inference, and the test checks the reported source."},{"criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"covered","evidence":["src/anomalyzer/seasonality.py:96","tests/test_analysis.py:247"],"notes":"The repeating seven-position series selects lag 7 and reports confirmation window [0, 27]."},{"criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"covered","evidence":["src/anomalyzer/seasonality.py:79","src/anomalyzer/seasonality.py:84","tests/test_analysis.py:247"],"notes":"Candidate-strength and early-confirmation failures return lag 1 with a reason; the quadratic example verifies fallback."},{"criterion":"The reserved calibration and evaluation tail does not determine positional season selection.","status":"covered","evidence":["src/anomalyzer/seasonality.py:69","tests/test_capability_behaviors.py:147"],"notes":"Discovery excludes calibration_size plus one reserved sample. Changing that tail preserves the complete inference diagnostics; arbitrary appended-prefix invariance is explicitly outside the guidance."}],"ignore_findings":[],"done":true,"input_fingerprint":"3317b1727dc3cabfcfc9b87808d9249d0d59f111b8fd5f8a1fad653b21343980","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Choose a repeating cycle"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","label":"Interpret rate changes wit...","area":"analysis","scopes":["analysis","analysis/rates"],"scope":"analysis/rates","scopeLabel":"analysis/rates","status":"planned","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","acceptance":["Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","Zero denominators remain unavailable, and low support is explicitly marked; intervals remain within the valid zero-to-one rate domain.","Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations."],"guidance":["Independence and stable exposure are modeling assumptions; clustered events, overdispersion, or changing populations require explicit limitations.","Proposed automated checks cover zero denominators, tiny counts, zero-or-all-success boundaries, and equal rates at different volumes.","Manual review should compare sparse and high-volume conversion changes and confirm the explanation discloses assumptions without inventing incident probability."],"path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Judge rates using their exposure"},"impact":0,"gaps":3,"r":47,"x":194,"y":524},{"id":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","label":"See changes before the day...","area":"analysis","scopes":["analysis","analysis/resolution"],"scope":"analysis/resolution","scopeLabel":"analysis/resolution","status":"implemented","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","acceptance":["Callers can select sum, mean, or last aggregation for completed and partial periods.","Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","An explicitly unfinished source sample is excluded from both analysis views.","A source cadence that does not evenly divide 24 hours is rejected."],"guidance":["Periods are fixed 24-hour windows, not daylight-saving-aware local calendar days.","Native and daily views share source observations and are correlated.","Legacy schema 1.0 requires complete input and does not support resets. Schema 1.1 labels sparse native/completed-day views and preserves reset boundaries; daily sums require declared additive volume and fully observed children."],"path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","dependencies":["data/quality/exclude-incomplete-periods","analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/sparse_views.py","tests/test_sparse.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"selected-aggregation","description":"Check sum, mean, and last for completed days and trailing partial aggregates.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_multi_resolution_uses_selected_aggregation"},{"id":"two-views","description":"Return completed days, native evidence, and a linked trailing partial-day marker.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_current_period_and_keeps_two_views"},{"id":"unfinished-source","description":"Exclude a marked incomplete source sample from both views.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days"},{"id":"dividing-cadence","description":"Reject subperiod durations that do not evenly divide a day.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_rejects_non_dividing_frequency"}],"manualChecks":["Inspect the partial aggregate's observed_aggregate and intraday_evidence_refs together with the two method namespaces."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","criteria":[{"criterion":"Callers can select sum, mean, or last aggregation for completed and partial periods.","status":"covered","evidence":["src/anomalyzer/contracts.py:18","src/anomalyzer/recipes.py:539","tests/test_capability_behaviors.py:178"],"notes":"Parameterized assertions check independently known completed and partial values for all three aggregation choices."},{"criterion":"Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","status":"covered","evidence":["src/anomalyzer/recipes.py:617","src/anomalyzer/recipes.py:667","src/anomalyzer/recipes.py:680","tests/test_analysis.py:170"],"notes":"Separate datasets, configurations, and namespaces preserve the two correlated views."},{"criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"covered","evidence":["src/anomalyzer/recipes.py:639","src/anomalyzer/recipes.py:690","tests/test_analysis.py:170"],"notes":"The marker contains finalized progress, observed aggregate, native evidence references, and an explicit exclusion flag."},{"criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"covered","evidence":["src/anomalyzer/validation.py:44","src/anomalyzer/recipes.py:564","tests/test_analysis.py:200"],"notes":"Both child datasets use prepared finalized observations; tests verify counts and retention of the excluded original value."},{"criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"covered","evidence":["src/anomalyzer/recipes.py:572","tests/test_analysis.py:221"],"notes":"A five-hour cadence raises the specific dividing-cadence error."}],"ignore_findings":[],"done":true,"input_fingerprint":"cdcdea1166a4e3d2523b594bec938142135e17a6b4279732e52819a486d1b4f2","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Compare finalized subperiods"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","label":"Compare days and weeks in ...","area":"data","scopes":["data","data/calendar"],"scope":"data/calendar","scopeLabel":"data/calendar","status":"planned","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","acceptance":["Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","Periods report UTC boundaries and local labels; daylight-saving days can contain 23 or 25 hours without being forced into fixed 24-hour windows.","Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation."],"guidance":["Calendar grouping needs a distinct contract from fixed elapsed cadence; it must not silently reinterpret existing frequency or aggregation settings.","Proposed automated checks cover spring transitions, fall transitions, week boundaries, and missing native slots in a calendar period.","Manual review should inspect a daylight-saving week and confirm local labels, observed coverage, and completed-versus-open comparisons are understandable."],"path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Declare local calendar boundaries"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","label":"Analyze many entities with...","area":"data","scopes":["data","data/cohorts"],"scope":"data/cohorts","scopeLabel":"data/cohorts","status":"planned","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","acceptance":["Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","Training, calibration, relationships, and cases do not cross entity boundaries unless a separate explicit aggregate comparison is requested.","The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident."],"guidance":["Batch orchestration should preserve existing per-request contracts rather than silently lifting schema limits; multiple-testing limits stay visible.","Proposed automated checks cover partition isolation, batch-versus-single equivalence, one-entity failure, and shared resource-budget exhaustion.","Manual review should inspect a mixed batch of healthy, changed, and sparse entities and confirm the summary does not equate missing support with health."],"path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Keep each entity's evidence separate"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","label":"Know when missing updates ...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"planned","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","acceptance":["Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","Results distinguish current, late, missing, and not-assessed freshness and expose the last observed update and the deadline used.","A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision."],"guidance":["Freshness depends on declared collection expectations; it does not prove source completeness and does not start a scheduler or send a notification.","Proposed automated checks cover exact deadlines, allowed lateness, absent source updates, and stale-source propagation to relationships.","Manual review should compare a stale feed with a true metric drop and confirm the report distinguishes collection uncertainty from numerical departure."],"path":".capabilities/data/quality/report-source-freshness.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Check whether the sources are current"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","label":"See how late or corrected ...","area":"data","scopes":["data","data/revisions"],"scope":"data/revisions","scopeLabel":"data/revisions","status":"planned","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","acceptance":["A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","Results preserve prior and new evidence versions and identify findings that were added, changed, or withdrawn with traceable source lineage.","Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results."],"guidance":["Schema-1.1 snapshot revision selection and as-of availability are implemented in data/quality/analyze-sparse-observations; durable prior/new evidence diffing and automatic reconciliation in this capability remain planned.","A corrected event-time result does not replace the record of what was known earlier; callers control retention and any notification about a revision.","Proposed automated checks cover late completion, corrected calibration data, downstream relationship changes, and conflicting source versions.","Manual review should inspect a drop withdrawn after ingestion completes and confirm both the original limitation and the revision explanation are retained."],"path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Reconcile revisions without losing history"},"impact":0,"gaps":3,"r":47,"x":353.33333333333337,"y":642},{"id":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","label":"Choose settings using held...","area":"evaluation","scopes":["evaluation","evaluation/comparisons"],"scope":"evaluation/comparisons","scopeLabel":"evaluation/comparisons","status":"planned","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","acceptance":["Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","The output preserves every candidate's settings and evaluation provenance; adopting a configuration requires an explicit caller decision."],"guidance":["Bounded comparison is not unrestricted parameter search; held-out results describe the supplied history and do not guarantee future effectiveness.","Proposed automated checks cover split isolation, comparable replay inputs, failed candidate retention, and unchanged active settings after comparison.","Manual review should compare an earlier but noisier configuration with a quieter one and confirm the tradeoff is visible without an unsupported winner."],"path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Review the tradeoff before adopting settings"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","label":"Measure detection timing a...","area":"evaluation","scopes":["evaluation","evaluation/outcomes"],"scope":"evaluation/outcomes","scopeLabel":"evaluation/outcomes","status":"planned","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","acceptance":["Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","Timing reports distinguish the incident's supplied start from the first available candidate, calibrated detection, and policy-eligible evidence.","Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown."],"guidance":["Detection timing does not measure response effectiveness or establish that an unmatched finding was false; incident labels are caller judgments.","Proposed automated checks cover boundary matching, ambiguous overlaps, incomplete label coverage, and separate candidate versus eligibility timing.","Manual review should inspect a matched and an unmatched episode and confirm the reported timing and accuracy limits agree with the supplied label coverage."],"path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Compare detection with reviewed outcomes"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","label":"Replay history as it would...","area":"evaluation","scopes":["evaluation","evaluation/replay"],"scope":"evaluation/replay","scopeLabel":"evaluation/replay","status":"planned","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","acceptance":["Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","Repeating a replay preserves numerical evidence and first-detection timing apart from generated IDs and runtime; stopped replays retain completed steps."],"guidance":["Event-time replay does not reconstruct arrival-time availability unless the caller supplies versioned arrival history.","Proposed automated checks cover future-tail isolation, short-history cutoffs, deterministic repeated replay, and interrupted replay retention.","Manual review should inspect one change from its first available observation through calibration and confirm the explanation uses only then-known evidence."],"path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Reconstruct the evidence available at the time"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","label":"Update evidence when the n...","area":"execution","scopes":["execution","execution/incremental"],"scope":"execution/incremental","scopeLabel":"execution/incremental","status":"planned","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","acceptance":["Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","Supported explicitly configured recipes produce the same numerical evidence as full batch analysis of the combined history, including cross-chunk patterns.","Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations."],"guidance":["Start with explicit season lengths; input-prefix season discovery is not currently guaranteed to be invariant when the history grows.","Proposed automated checks cover chunk-versus-batch equivalence, cross-chunk CUSUM, idempotent retry, and failed-append state preservation.","Manual review should follow a sustained change across several append batches and confirm detection timing and provisional evidence remain understandable."],"path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Analyze the next finalized samples"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","label":"Keep useful evidence when ...","area":"execution","scopes":["execution","execution/resilience"],"scope":"execution/resilience","scopeLabel":"execution/resilience","status":"implemented","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","acceptance":["A method failure preserves evidence completed before the failure and reports the error.","Runtime budgets are checked cooperatively and produce an explicit stop reason.","CLI byte limits and execution failures produce distinct error exits.","A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial."],"guidance":["Default limits are 10,000 source observations per request, 10 MB CLI input, and 60 seconds cooperative runtime.","Numerical results are deterministic for fixed declared settings; generated run IDs and runtimes vary."],"path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/budget.py","src/anomalyzer/recipes.py","src/anomalyzer/orchestrator.py","src/anomalyzer/io.py","src/anomalyzer/cli.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"method-and-budget","description":"Retain pre-failure evidence and report a runtime-budget stop.","command":"python -m pytest -q tests/test_analysis.py::test_failure_and_budget"},{"id":"cli-limits","description":"Enforce byte limits and distinguish execution failure in CLI output and exit status.","command":"python -m pytest -q tests/test_cli.py::test_size_limit_and_execution_exit"},{"id":"partial-siblings","description":"Preserve healthy analysis when a sibling dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Inspect run status, stop_reason, and each target's status before relying on a result; confirm limits are presented as cooperative rather than hard process timeouts."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","criteria":[{"criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"covered","evidence":["src/anomalyzer/recipes.py:400","src/anomalyzer/recipes.py:448","tests/test_analysis.py:509"],"notes":"The failure regression preserves evidence through index 44 and reports the injected error at index 45."},{"criterion":"Runtime budgets are checked cooperatively and produce an explicit stop reason.","status":"covered","evidence":["src/anomalyzer/budget.py:18","src/anomalyzer/recipes.py:108","src/anomalyzer/recipes.py:508","tests/test_analysis.py:509"],"notes":"Cooperative target/sample checks expose runtime_budget and budget_exceeded; this is not a hard process timeout."},{"criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:429","tests/test_cli.py:108"],"notes":"The full suite executed the byte-limit and execution-exit regression successfully; invalid input exits 2 and execution failure exits 1."},{"criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:125","src/anomalyzer/orchestrator.py:171","src/anomalyzer/orchestrator.py:185","tests/test_capability_behaviors.py:159","tests/test_capability_behaviors.py:30"],"notes":"Healthy dataset evidence survives target-specific failure; inspected incomplete-usage reports partial with no invented healthy relationship."}],"ignore_findings":[],"done":true,"input_fingerprint":"146a9c8396fd7a2d039f75077dc0b82a0778685d20c3782aa284d5f0b6ed94be","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect incomplete execution"},"impact":0,"gaps":0,"r":38,"x":720,"y":642},{"id":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","label":"Resume local analysis afte...","area":"execution","scopes":["execution","execution/state"],"scope":"execution/state","scopeLabel":"execution/state","status":"planned","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","acceptance":["A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","Restoring a supported state and continuing yields the same numerical evidence and detection positions as uninterrupted processing of the same observations.","Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled."],"guidance":["Stored observations and context can contain sensitive data; document exactly what a snapshot contains and avoid automatic persistence.","Proposed automated checks cover restart equivalence, corrupt snapshots, version or settings mismatch, and output overwrite protection.","Manual review should inspect a snapshot and resumed explanation and confirm their provenance and retained-data scope are clear to an application owner."],"path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Preserve state across restarts"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","label":"Understand findings in a r...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","acceptance":["Point-anomaly episode counts and residual-pattern findings are labeled separately.","Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","Readiness status remains visible, including when it was not assessed."],"guidance":["Readable summaries should retain the evidence's claim limits and allow drilldown into JSON."],"path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/cli.py","tests/test_cli.py","examples/agentic/conversion.json","examples/agentic/unit-cost.json"],"automatedChecks":[{"id":"episodes-patterns","description":"Distinguish shift patterns from point episodes in readable output.","command":"python -m pytest -q tests/test_cli.py::test_readable_output_distinguishes_episodes_from_patterns"},{"id":"relationship-case","description":"Explain a conversion case with entity, units, comparison, and a suggested check.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_relationship_request_through_cli"},{"id":"explicit-rule","description":"Describe the unit-cost rule boundary and unassessed readiness.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the conversion and unit-cost example output and confirm the descriptions state the supported numerical comparison in plain language."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","criteria":[{"criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"covered","evidence":["src/anomalyzer/cli.py:347","src/anomalyzer/cli.py:405","tests/test_cli.py:143"],"notes":"Readable tests distinguish zero-episode residual shifts from a spike episode and retain the overlap warning."},{"criterion":"Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","status":"covered","evidence":["src/anomalyzer/cli.py:291","src/anomalyzer/cli.py:308","src/anomalyzer/cli.py:331","tests/test_cli.py:117","examples/agentic/conversion.json"],"notes":"Conversion output identifies site=demo, observed/expected orders_per_visit, the relationship, and a collection-completeness investigation question."},{"criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"covered","evidence":["src/anomalyzer/cli.py:318","tests/test_cli.py:134","examples/agentic/unit-cost.json"],"notes":"Unit-cost output names review-above-13c and its [0, 0.13] boundary as a declared rule crossing."},{"criterion":"Readiness status remains visible, including when it was not assessed.","status":"covered","evidence":["src/anomalyzer/cli.py:352","src/anomalyzer/cli.py:399","tests/test_cli.py:134"],"notes":"Readable output includes not_assessed and its supporting reasons without claiming measured alert accuracy."}],"ignore_findings":[],"done":true,"input_fingerprint":"98108868b000972fd9b3c3f082846fbd3f3b08c399ef7ed424c23245d6935fef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Read the findings"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","label":"Share an investigation tha...","area":"interfaces","scopes":["interfaces","interfaces/export"],"scope":"interfaces/export","scopeLabel":"interfaces/export","status":"planned","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","acceptance":["Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","Source observations and context are included only by explicit selection; omitted data and resulting limits on independent reproduction are disclosed."],"guidance":["Fingerprints identify included provenance; they do not prove source truth. A portable report must make privacy scope and reproduction limits inspectable.","Proposed automated checks cover manifest consistency, partial-run export, omitted-source disclosure, and overwrite protection.","Manual review should open a bundle outside the original session and confirm a reviewer can trace one finding and see exactly what cannot be reproduced."],"path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Preserve a portable evidence record"},"impact":0,"gaps":3,"r":47,"x":927.3333333333334,"y":760},{"id":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","label":"Install a versioned analys...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","acceptance":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"guidance":["Guard owns scheduling, persistence, business severity, incidents and notifications.","A published package does not establish integration into UsageTap Guard."],"path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","interfaces/cli/analyze-files-and-streams","interfaces/mcp/connect-read-only-local-server"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["pyproject.toml","src/anomalyzer/_version.py","scripts/release.py","scripts/publish_release.py","scripts/verify_distribution.py","scripts/verify_lambda.sh","examples/library/guard_smoke.py",".github/workflows/ci.yml",".github/workflows/release.yml","docs/releases/README.md","tests/test_release.py"],"automatedChecks":[{"id":"release-safety","description":"Reject mismatched tags, altered assets and mutable publication.","command":"python -m pytest -q tests/test_release.py"},{"id":"installed-artifacts","description":"Install wheel and sdist outside the checkout and invoke public APIs.","command":"python scripts/verify_distribution.py --python 3.12"},{"id":"release-inventory","description":"Check archive contents, version metadata and SHA-256 inventory.","command":"python scripts/release.py verify"},{"id":"linux-offline-consumer","description":"Bundle and run with networking disabled in Linux Python 3.12.","command":"bash scripts/verify_lambda.sh"}],"manualChecks":["Follow the release guide and confirm Guard retains domain severity and incident ownership; inspect readiness and evidence limits before adopting."],"verificationGaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for interfaces/python/install-versioned-library.","criteria":[{"criterion":"Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","status":"partial","evidence":["src/anomalyzer/_version.py:1","scripts/release.py:1","scripts/release.py:20","scripts/release.py:39","scripts/release.py:48","scripts/release.py:54","scripts/publish_release.py:19","scripts/publish_release.py:27","scripts/publish_release.py:29","scripts/publish_release.py:39","scripts/verify_lambda.sh:14","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19","scripts/verify_lambda.sh:25","scripts/verify_lambda.sh:27",".github/workflows/ci.yml:14",".github/workflows/ci.yml:19",".github/workflows/ci.yml:27",".github/workflows/ci.yml:41",".github/workflows/ci.yml:46",".github/workflows/release.yml:28","docs/releases/README.md:12","docs/releases/README.md:13","docs/releases/README.md:14","docs/releases/README.md:17","docs/releases/README.md:18","tests/test_release.py:60","tests/test_release.py:71","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","status":"partial","evidence":["scripts/verify_distribution.py:1","scripts/verify_distribution.py:42","examples/library/guard_smoke.py:32","examples/library/guard_smoke.py:49","examples/library/guard_smoke.py:65","examples/library/guard_smoke.py:69","examples/library/guard_smoke.py:75",".github/workflows/ci.yml:57","docs/releases/README.md:13","docs/releases/README.md:208","docs/releases/README.md:211","docs/releases/README.md:213"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:34","scripts/release.py:81","scripts/release.py:82","scripts/release.py:85","scripts/release.py:92","scripts/publish_release.py:1","scripts/publish_release.py:25","scripts/publish_release.py:26","scripts/publish_release.py:27","scripts/publish_release.py:28","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:24","examples/library/guard_smoke.py:39",".github/workflows/ci.yml:42",".github/workflows/ci.yml:45",".github/workflows/ci.yml:46",".github/workflows/ci.yml:49",".github/workflows/ci.yml:51",".github/workflows/release.yml:31",".github/workflows/release.yml:33",".github/workflows/release.yml:36",".github/workflows/release.yml:37","docs/releases/README.md:3","docs/releases/README.md:12","docs/releases/README.md:14","docs/releases/README.md:15","docs/releases/README.md:19","tests/test_release.py:1","tests/test_release.py:21","tests/test_release.py:23","tests/test_release.py:52","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:81","scripts/release.py:82","scripts/publish_release.py:40","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:6","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19",".github/workflows/ci.yml:30",".github/workflows/ci.yml:31",".github/workflows/ci.yml:51",".github/workflows/ci.yml:52",".github/workflows/ci.yml:90","docs/releases/README.md:12","docs/releases/README.md:15","docs/releases/README.md:58","docs/releases/README.md:59","docs/releases/README.md:75"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","partial: Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","partial: Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","partial: A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","name":"Python integration","label":"Python integration / Embed analysis in my application / Install a released dependency"},"impact":0,"gaps":1,"r":41,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","label":"Ask an agent to explain un...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","acceptance":["The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit."],"guidance":["This contract covers the supplied instructions and demonstrated analysis workflow; it does not guarantee that every agent host follows the instructions.","Synthetic examples demonstrate mechanics, not real-world detection accuracy."],"path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/agents/openai.yaml","src/anomalyzer/cli.py","src/anomalyzer/mcp.py","tests/test_cli.py","tests/test_agentic.py","tests/test_mcp.py"],"automatedChecks":[{"id":"structured-cli","description":"Exercise the documented structured CLI workflow and preserve incomplete-history evidence.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence tests/test_cli.py::test_bare_json_values_cli"},{"id":"declared-relationships","description":"Check explicit ratio semantics and unavailable source evidence.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"mcp-drilldown","description":"Follow compact analysis to a finding and its source evidence.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case"}],"manualChecks":["Follow the conversion example through the CLI and MCP drilldown; compare the Skill instructions with actual evidence and check that its proposed text separates calibrated findings, early candidates, readiness, and caller-owned actions."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","criteria":[{"criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:13",".agents/skills/anomaly-detection/SKILL.md:41","tests/test_cli.py:13"],"notes":"Executable resolution and JSON commands are explicit. Manual conversion and spike CLI runs use the installed numerical engine."},{"criterion":"The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:24",".agents/skills/anomaly-detection/SKILL.md:32","tests/test_cli.py:51","tests/test_agentic.py:211"],"notes":"Instructions preserve scope, units, missingness, declared relationships, and exact alignment; existing runtime checks establish these prerequisites."},{"criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:101",".agents/skills/anomaly-detection/SKILL.md:134",".agents/skills/anomaly-detection/agents/openai.yaml:4"],"notes":"The supplied instructions lead with plain-language changes and then qualify maturity/readiness/overlap. This establishes instructional content, not measured agent compliance."},{"criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:54",".agents/skills/anomaly-detection/SKILL.md:174","src/anomalyzer/mcp.py:169","tests/test_mcp.py:54"],"notes":"The documented three-tool workflow matches executed conversion drilldown, with session-only run IDs and explicit read-only/caller-owned delivery limits."}],"ignore_findings":[],"done":true,"input_fingerprint":"751723214aa340b1227ec347c22f130749fcdbfd49327c21b4c27ef68d7f0341","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Get an evidence-grounded explanation"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","label":"See the evidence in a char...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","acceptance":["The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes."],"guidance":["Native chart rendering and interactive figure features depend on the agent host; the checked-in fallback produces a static SVG for a single dataset.","A chart explains structured analysis; it does not replace the detector result."],"path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/scripts/quick_chart.py",".agents/skills/anomaly-detection/references/visualization.md","tests/test_skill_chart.py"],"automatedChecks":[{"id":"observed-evidence","description":"Render real observed and expected data and only actual point triggers.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_preserves_real_point_evidence"},{"id":"maturity-and-resets","description":"Keep provisional evidence non-triggering and label reviewed regimes.","command":"python -m pytest -q tests/test_skill_chart.py::test_provisional_scores_are_not_charted_as_triggers tests/test_skill_chart.py::test_quick_chart_marks_reviewed_resets"},{"id":"unsupported-scope","description":"Reject a multi-dataset result rather than mixing units.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_rejects_multiple_datasets"},{"id":"output-protection","description":"Require an explicit overwrite for existing charts.","command":"python -m pytest -q tests/test_skill_chart.py::test_chart_cli_requires_explicit_overwrite"}],"manualChecks":["Inspect a spike fallback chart and the detailed figure instructions; verify that axes, initialization context, trigger markers, and interpretation limits are clear, and that the Skill requests a chart only when the user asks."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","criteria":[{"criterion":"The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:189",".agents/skills/anomaly-detection/SKILL.md:226"],"notes":"The Skill explicitly requires a visual request after analysis and separates quick versus detailed rendering instructions."},{"criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:81",".agents/skills/anomaly-detection/scripts/quick_chart.py:59","tests/test_skill_chart.py:28","tests/test_skill_chart.py:43","tests/test_skill_chart.py:54"],"notes":"Real-result tests establish observed/expected lines, trigger-only markers, provisional non-triggering, units, initialization shading, and reset boundaries; the spike chart was inspected visually."},{"criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:85",".agents/skills/anomaly-detection/scripts/quick_chart.py:251","tests/test_skill_chart.py:63","tests/test_skill_chart.py:68"],"notes":"Multiple resolved datasets raise a specific error. The CLI test preserves an existing chart and replaces it only with explicit overwrite."},{"criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/references/visualization.md:31",".agents/skills/anomaly-detection/references/visualization.md:37",".agents/skills/anomaly-detection/references/visualization.md:45",".agents/skills/anomaly-detection/references/visualization.md:86"],"notes":"Detailed instructions require index alignment, maturity distinctions, first detection timing, visible unavailable status, and qualified causal claims; native interactive rendering remains host-dependent."}],"ignore_findings":[],"done":true,"input_fingerprint":"2780b6c3e363f89a30a0709772c0a46a9a96f1567fabb27e63934a3ca80dc982","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Inspect a requested chart"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","label":"Find which segments contri...","area":"investigation","scopes":["investigation","investigation/cohorts"],"scope":"investigation/cohorts","scopeLabel":"investigation/cohorts","status":"planned","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","acceptance":["Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","Segment ranking retains absolute signed contributions and baseline support; an arithmetic contributor is not automatically labeled anomalous or causal."],"guidance":["Ratios and overlapping customer groups are not additive totals; separate segment expectations must not be assumed to sum to a fitted total expectation.","Proposed automated checks cover exact reconciliation, offsetting segment changes, missing segments, and zero total departure.","Manual review should inspect a regional decline with an offsetting increase and confirm the explanation preserves signed contributions and evidence limits."],"path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Locate contributors to a total change"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","label":"Keep reviewed outcomes bes...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","acceptance":["A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","Review export distinguishes analyst judgment from detector output; annotation alone does not change calibration, eligibility, or regime boundaries."],"guidance":["Caller-supplied reviewer identity is attribution, not built-in authentication; storage, access control, response execution, and incident ownership stay with callers.","Proposed automated checks cover evidence-version linking, unresolved reviews, revision retention, and unchanged numerical evidence after annotation.","Manual review should inspect a disputed case with two review revisions and confirm judgments, uncertainties, and original findings remain distinguishable."],"path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Record the reviewed outcome"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","label":"Find earlier reviewed case...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","acceptance":["Search is bounded to caller-selected records and explicit compatibility filters; it does not scan unrelated local files or require a model API.","Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match."],"guidance":["Similar historical evidence does not establish the same cause or authorize repeating a past response; caller-selected records may be incomplete or biased.","Proposed automated checks cover compatibility filtering, bounded corpus access, missing comparison context, and no-match behavior.","Manual review should compare superficially similar cases with different outcomes and confirm precedent is presented as investigation context."],"path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Inspect comparable reviewed history"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","label":"Change action eligibility ...","area":"investigation","scopes":["investigation","investigation/policy"],"scope":"investigation/policy","scopeLabel":"investigation/policy","status":"implemented","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","acceptance":["Default policy leaves action and notification eligibility disabled.","Explicit-rule and calibrated-departure eligibility are controlled independently.","Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged."],"guidance":["Eligibility does not send notifications, execute actions, check readiness, or confer authority to act.","Monitoring schedules, delivery, reviewed incident state, and deduplication belong to the caller."],"path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","dependencies":["investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/agent.py","src/anomalyzer/cases.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"independent-notifications","description":"Replay notification and action choices without changing numerical evidence or the original result.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_separates_action_from_notification"},{"id":"calibrated-choice","description":"Enable calibrated departures only with the corresponding policy choice.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_enables_calibrated_departures_only_when_requested"},{"id":"nontriggering-candidates","description":"Keep descriptive candidates ineligible even under a permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"}],"manualChecks":["Inspect the unit-cost example under default and allow_explicit_rules policies and compare eligibility with the unchanged numerical violation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","criteria":[{"criterion":"Default policy leaves action and notification eligibility disabled.","status":"covered","evidence":["src/anomalyzer/contracts.py:329","tests/test_capability_behaviors.py:74"],"notes":"Default flags disable both eligibility types, including after replay resets policy."},{"criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"covered","evidence":["src/anomalyzer/agent.py:47","tests/test_capability_behaviors.py:123"],"notes":"Explicit-only permission does not enable a calibrated departure; the matching calibrated choice does."},{"criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"covered","evidence":["src/anomalyzer/agent.py:53","tests/test_capability_behaviors.py:101","tests/test_capability_behaviors.py:134"],"notes":"Notification-only replay enables a met rule while keeping action disabled; permissive settings still cannot enable candidates."},{"criterion":"Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged.","status":"covered","evidence":["src/anomalyzer/agent.py:42","src/anomalyzer/cases.py:57","src/anomalyzer/cases.py:105","tests/test_capability_behaviors.py:101","tests/test_agentic.py:49"],"notes":"Assertions compare unchanged original serialization, methods, evidence references, stable IDs, and eligibility-dependent revision changes."}],"ignore_findings":[],"done":true,"input_fingerprint":"ee67bcc1ec1380b8d84a0556d838d648ae6cef3f4f9b9e34e1b71e90eb00aeae","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Review action eligibility"},"impact":0,"gaps":0,"r":38,"x":1086.6666666666667,"y":878},{"id":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","label":"Review a possible new base...","area":"investigation","scopes":["investigation","investigation/regimes"],"scope":"investigation/regimes","scopeLabel":"investigation/regimes","status":"planned","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","acceptance":["A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","Producing a proposal leaves the active regime and original findings unchanged.","A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect."],"guidance":["A proposed boundary is not a confirmed cause or a retrospectively available detection; adaptive reference updates remain a separate behavior.","Proposed automated checks cover unchanged active results, short new regimes, rejected proposals, and accepted explicit-reset equivalence.","Manual review should inspect a persistent outage-like drop and a known launch and confirm neither becomes ordinary behavior merely because it persists."],"path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Review a proposed regime transition"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","label":"Distinguish changing perfo...","area":"relationships","scopes":["relationships","relationships/cohorts"],"scope":"relationships/cohorts","scopeLabel":"relationships/cohorts","status":"planned","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","acceptance":["Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","The report preserves aggregate and cohort rates, population weights, and the declared decomposition formula and reference-period convention.","Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes."],"guidance":["Decomposition is descriptive accounting; it does not estimate an intervention effect or establish that the comparison populations are causally equivalent.","Proposed automated checks cover stable rates with changing mix, stable mix with changing rates, offsetting cohort rates, and new or empty cohorts.","Manual review should inspect a Simpson's-paradox example and confirm the aggregate movement and cohort evidence can both be understood correctly."],"path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Separate performance changes from population mix"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878},{"id":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","label":"Check whether a metric fol...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","acceptance":["A supplied lag and coefficient use the predictor at the declared earlier sample position.","Derived values retain response and predictor source indexes and source cutoffs in lineage.","When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","Samples lacking the earlier predictor or awaiting coefficient training remain unavailable."],"guidance":["Lag is measured in samples and is never searched automatically.","Exactly one of coefficient or training_size is required."],"path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","dependencies":["relationships/quality/require-compatible-sources"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_examples.py","examples/agentic/lagged-response.json"],"automatedChecks":[{"id":"supplied-response","description":"Use the declared lag and coefficient with correct predictor indexes.","command":"python -m pytest -q tests/test_agentic.py::test_lagged_response_uses_supplied_coefficient_and_prior_sample"},{"id":"frozen-coefficient","description":"Keep fitted coefficient and earlier lineage invariant when future samples arrive.","command":"python -m pytest -q tests/test_agentic.py::test_estimated_lag_coefficient_is_frozen_and_prefix_invariant"},{"id":"delayed-example","description":"Run the checked-in lagged-response demonstration against expected results.","command":"python -m pytest -q tests/test_examples.py::test_agentic_demo_checks_all_cases"},{"id":"pending-training","description":"Retain unavailable lineage until the full declared training prefix arrives, then freeze its coefficient.","command":"python -m pytest -q tests/test_agentic.py::test_lag_fit_waits_for_complete_training_prefix"}],"manualChecks":["Inspect the lagged-response example's leading unavailable values, coefficient parameters, and first usable derived sample."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","criteria":[{"criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"covered","evidence":["src/anomalyzer/relationships.py:115","tests/test_agentic.py:254"],"notes":"Supplied lag/coefficient uses the declared earlier predictor position and preserves the expected response residual."},{"criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"covered","evidence":["src/anomalyzer/relationships.py:152","tests/test_agentic.py:254","examples/agentic/lagged-response.json"],"notes":"Available and pending lineage retains existing source indexes and timestamp cutoffs; missing predictors do not become zero."},{"criterion":"When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","status":"covered","evidence":["src/anomalyzer/relationships.py:47","tests/test_agentic.py:294","tests/test_agentic.py:330"],"notes":"The coefficient is fitted once after the complete declared prefix; the independent 20/7 result and prefix-invariance assertions pass."},{"criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"covered","evidence":["src/anomalyzer/relationships.py:122","tests/test_agentic.py:330"],"notes":"Short prefixes retain null derived values and explicit unavailable reasons, while an unidentifiable completed fit has a visible error."}],"ignore_findings":[],"done":true,"input_fingerprint":"fc4516ef6ab903dfa4918268d97049aff04fbf9556620ccb8259ff830ba85785","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Check delayed responses"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","label":"Inspect related metrics be...","area":"relationships","scopes":["relationships","relationships/resolution"],"scope":"relationships/resolution","scopeLabel":"relationships/resolution","status":"implemented","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed fixed UTC 24-hour additive aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","acceptance":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"guidance":["Daily ratios and differences are supported. Daily lagged, normalized-residual and joint-condition definitions remain explicitly inapplicable; their native evidence remains eligible. Local-calendar aggregation is separate work.","Initial support should use explicit relationships and compatible boundaries; cross-cadence joining or automatic resampling needs a separate contract.","Proposed automated checks cover ratio-of-totals arithmetic, missing sources, mismatched period boundaries, and open-period evidence separation.","Manual review should inspect a conversion decline within an unfinished day and verify the explanation separates native evidence from partial totals."],"path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","dependencies":["data/quality/analyze-sparse-observations","relationships/comparisons/compare-declared-metrics"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/sparse_views.py","src/anomalyzer/relationships.py","src/anomalyzer/agent.py","tests/test_sparse.py","examples/sparse/unfinished_daily.json","docs/sparse-data.md"],"automatedChecks":[{"id":"ratio-of-sums","description":"Compute daily ratios from compatible sums and preserve missing-child coverage.","command":"python -m pytest -q tests/test_sparse.py::test_complete_daily_ratio_divides_sums_and_missing_child_invalidates_only_its_view"},{"id":"view-policies","description":"Keep native/daily rules separate and preserve case identity on replay.","command":"python -m pytest -q tests/test_sparse.py::test_daily_rules_are_not_silently_reused_from_hourly_rules tests/test_sparse.py::test_multi_resolution_mcp_drilldown_selects_correct_view_and_replays_case_ids"},{"id":"view-limits","description":"Report unsupported daily lag semantics without discarding native evidence.","command":"python -m pytest -q tests/test_sparse.py::test_unsupported_daily_relationship_is_explicit_and_does_not_reinterpret_lag"}],"manualChecks":["Inspect progression chapter 22 and confirm incomplete progress is not presented as a completed total or independent corroboration."],"verificationGaps":[],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for relationships/resolution/compare-finalized-subperiods.","criteria":[{"criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:19","src/anomalyzer/sparse_views.py:22","src/anomalyzer/sparse_views.py:139","src/anomalyzer/sparse_views.py:143","src/anomalyzer/sparse_views.py:164","src/anomalyzer/agent.py:14","src/anomalyzer/agent.py:20","src/anomalyzer/agent.py:23","src/anomalyzer/agent.py:29","tests/test_sparse.py:18","tests/test_sparse.py:59","tests/test_sparse.py:87","tests/test_sparse.py:99","tests/test_sparse.py:117","docs/sparse-data.md:31","docs/sparse-data.md:127","docs/sparse-data.md:159","docs/sparse-data.md:199","docs/sparse-data.md:203"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:19","src/anomalyzer/sparse_views.py:75","src/anomalyzer/sparse_views.py:94","src/anomalyzer/sparse_views.py:122","src/anomalyzer/sparse_views.py:123","src/anomalyzer/relationships.py:7","src/anomalyzer/relationships.py:19","src/anomalyzer/relationships.py:33","src/anomalyzer/relationships.py:34","src/anomalyzer/relationships.py:36","src/anomalyzer/agent.py:34","tests/test_sparse.py:10","tests/test_sparse.py:40","tests/test_sparse.py:50","tests/test_sparse.py:60","tests/test_sparse.py:62","docs/sparse-data.md:17","docs/sparse-data.md:41","docs/sparse-data.md:68","docs/sparse-data.md:97","docs/sparse-data.md:99"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:70","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:170","src/anomalyzer/relationships.py:90","src/anomalyzer/relationships.py:173","tests/test_sparse.py:230","tests/test_sparse.py:259","tests/test_sparse.py:313","docs/sparse-data.md:41","docs/sparse-data.md:111","docs/sparse-data.md:194","docs/sparse-data.md:197","docs/sparse-data.md:198"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:1","src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:130","src/anomalyzer/sparse_views.py:159","src/anomalyzer/relationships.py:157","src/anomalyzer/relationships.py:192","src/anomalyzer/agent.py:20","src/anomalyzer/agent.py:47","tests/test_sparse.py:217","tests/test_sparse.py:247","tests/test_sparse.py:251","tests/test_sparse.py:264","tests/test_sparse.py:386","examples/sparse/unfinished_daily.json:208","docs/sparse-data.md:8","docs/sparse-data.md:11","docs/sparse-data.md:37","docs/sparse-data.md:41","docs/sparse-data.md:45"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","partial: Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","partial: Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","partial: Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"reviewStale":false,"health":"gap","healthLabel":"Review findings","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","name":"Related-metric analysis","label":"Related-metric analysis / Monitor the periods my business uses / Compare related metrics while the period is open"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878}],"links":[{"source":"data/inputs/accept-analysis-requests","target":"analysis/evidence/calibrate-departures","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/cli/analyze-files-and-streams","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"analysis/expectations/model-seasonal-level-and-growth","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"interfaces/python/analyze-through-public-api","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/identify-residual-patterns","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"investigation/assessments/describe-early-changes","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"relationships/quality/require-compatible-sources","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"analysis/expectations/reset-declared-regimes","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"data/quality/exclude-incomplete-periods","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/mcp/connect-read-only-local-server","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/rules/check-declared-numerical-rules","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/report-point-anomaly-episodes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/evidence/assess-reference-readiness","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/adapt-to-changing-growth","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/protect-reference-from-extremes","kind":"committed"},{"source":"data/quality/exclude-incomplete-periods","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"execution/resilience/retain-evidence-under-failure-and-limits","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/mcp/connect-read-only-local-server","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/visualize-analysis-on-request","kind":"committed"},{"source":"investigation/rules/check-declared-numerical-rules","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/evaluate-delayed-response","kind":"committed"},{"source":"data/quality/analyze-sparse-observations","target":"relationships/resolution/compare-finalized-subperiods","kind":"committed"},{"source":"relationships/comparisons/compare-declared-metrics","target":"relationships/resolution/compare-finalized-subperiods","kind":"committed"}]};
 const width = 1440;
 const height = 980;
 const padding = 74;
diff --git a/.capabilities/dependency-preview.svg b/.capabilities/dependency-preview.svg
index 9198df1..adf0e31 100644
--- a/.capabilities/dependency-preview.svg
+++ b/.capabilities/dependency-preview.svg
@@ -81,45 +81,50 @@
   </g>
   <g id="graph">
     <g id="links">
-      <path class="link committed" d="M728.5,235.5 Q500.6,375.3 412.9,627.8" />
-      <path class="link committed" d="M912.3,209.5 Q812.5,222.9 781.9,318.9" />
-      <path class="link committed" d="M702.4,179.9 Q448.1,86.0 191.0,172.0" />
-      <path class="link committed" d="M842.3,177.2 Q870.1,209.3 894.4,174.4" />
-      <path class="link committed" d="M1276.6,284.8 Q1262.6,218.0 1194.6,213.1" />
-      <path class="link committed" d="M166.6,299.4 Q636.5,328.6 1082.9,179.3" />
-      <path class="link committed" d="M973.5,229.0 Q944.2,331.6 1020.1,406.5" />
-      <path class="link committed" d="M1125.1,227.1 Q1037.9,297.7 1056.1,408.4" />
-      <path class="link committed" d="M840.2,198.4 Q1028.2,339.5 1260.9,306.3" />
-      <path class="link committed" d="M347.3,209.2 Q714.7,384.6 1121.7,384.3" />
-      <path class="link committed" d="M322.9,626.3 Q303.6,444.4 146.2,351.3" />
-      <path class="link committed" d="M347.1,210.1 Q436.9,291.5 554.0,260.2" />
-      <path class="link committed" d="M995.3,425.6 Q989.3,372.2 935.8,368.7" />
-      <path class="link committed" d="M1263.3,315.7 Q1191.7,261.5 1111.1,301.1" />
-      <path class="link committed" d="M302.3,732.8 Q195.0,741.0 151.9,839.6" />
-      <path class="link committed" d="M393.3,620.6 Q567.1,446.6 575.8,200.9" />
-      <path class="link committed" d="M309.2,637.5 Q271.6,515.0 147.4,482.9" />
-      <path class="link committed" d="M125.5,231.0 Q29.4,472.0 113.2,717.6" />
-      <path class="link committed" d="M186.9,168.2 Q293.9,220.4 395.6,158.3" />
-      <path class="link committed" d="M374.7,615.2 Q494.9,423.4 437.1,204.4" />
-      <path class="link committed" d="M571.9,309.1 Q350.1,411.1 258.4,637.4" />
-      <path class="link committed" d="M138.4,229.9 Q88.5,444.5 215.2,624.8" />
-      <path class="link committed" d="M745.4,245.6 Q566.7,439.2 554.0,702.3" />
-      <path class="link committed" d="M755.8,422.4 Q749.8,591.2 880.5,698.1" />
-      <path class="link committed" d="M1122.7,226.1 Q938.7,426.4 920.0,697.7" />
-      <path class="link committed" d="M931.1,224.5 Q737.1,407.5 703.7,672.1" />
-      <path class="link committed" d="M720.1,427.6 Q639.7,538.5 687.2,667.0" />
-      <path class="link committed" d="M872.3,375.1 Q719.2,485.1 705.5,673.1" />
-      <path class="link committed" d="M747.8,425.4 Q723.5,520.8 796.6,586.7" />
-      <path class="link committed" d="M991.3,496.8 Q890.0,510.0 849.8,604.0" />
-      <path class="link committed" d="M773.5,410.6 Q804.2,529.9 921.4,567.9" />
-      <path class="link committed" d="M1071.3,344.0 Q1042.7,529.1 1165.9,670.3" />
-      <path class="link committed" d="M1150.0,230.9 Q1073.5,455.1 1176.6,668.4" />
-      <path class="link committed" d="M1157.4,447.8 Q1053.3,533.0 1062.6,667.2" />
-      <path class="link committed" d="M438.4,691.9 Q713.2,791.1 994.1,710.5" />
-      <path class="link committed" d="M1197.4,449.3 Q1178.2,639.3 1311.8,775.9" />
+      <path class="link committed" d="M726.7,209.9 Q538.3,179.7 406.3,317.4" />
+      <path class="link committed" d="M941.8,225.6 Q906.2,230.9 925.7,261.1" />
+      <path class="link committed" d="M722.8,183.0 Q606.9,110.6 485.5,173.5" />
+      <path class="link committed" d="M872.5,179.4 Q890.4,208.4 902.8,176.7" />
+      <path class="link committed" d="M1203.1,236.2 Q1215.2,211.3 1188.8,219.5" />
+      <path class="link committed" d="M337.7,166.3 Q709.8,258.5 1083.0,170.6" />
+      <path class="link committed" d="M311.4,270.2 Q336.7,237.8 298.4,222.9" />
+      <path class="link committed" d="M1005.2,216.9 Q998.5,257.0 1039.1,257.3" />
+      <path class="link committed" d="M1118.9,224.0 Q1088.2,224.6 1104.8,250.3" />
+      <path class="link committed" d="M871.2,201.3 Q1009.6,313.0 1181.2,265.9" />
+      <path class="link committed" d="M279.2,624.8 Q808.0,489.6 1264.0,189.8" />
+      <path class="link committed" d="M361.6,169.0 Q268.1,111.1 171.0,162.8" />
+      <path class="link committed" d="M274.1,294.1 Q248.8,206.4 158.2,195.9" />
+      <path class="link committed" d="M723.5,175.8 Q700.8,139.3 669.1,168.3" />
+      <path class="link committed" d="M167.0,161.0 Q364.3,250.7 561.6,161.0" />
+      <path class="link committed" d="M337.7,165.2 Q450.5,223.4 561.6,161.8" />
+      <path class="link committed" d="M273.5,614.9 Q512.8,553.5 646.9,346.0" />
+      <path class="link committed" d="M1063.2,355.6 Q1028.9,372.9 1050.3,404.7" />
+      <path class="link committed" d="M1219.3,329.3 Q1179.7,346.9 1197.4,386.5" />
+      <path class="link committed" d="M1319.2,221.0 Q1277.6,286.5 1321.9,350.2" />
+      <path class="link committed" d="M406.2,347.2 Q832.1,463.8 1269.1,400.8" />
+      <path class="link committed" d="M302.5,411.7 Q133.6,582.6 130.9,822.8" />
+      <path class="link committed" d="M264.3,308.9 Q259.2,273.1 227.5,290.4" />
+      <path class="link committed" d="M279.6,396.8 Q171.3,427.6 145.8,537.2" />
+      <path class="link committed" d="M466.4,210.8 Q464.6,245.9 499.2,239.8" />
+      <path class="link committed" d="M644.4,333.0 Q426.3,324.1 274.9,481.2" />
+      <path class="link committed" d="M391.2,222.8 Q263.6,305.4 254.2,457.1" />
+      <path class="link committed" d="M819.7,257.7 Q793.4,362.6 873.7,435.0" />
+      <path class="link committed" d="M914.7,386.6 Q888.6,479.4 959.2,545.2" />
+      <path class="link committed" d="M1124.1,226.4 Q981.8,355.3 991.2,547.1" />
+      <path class="link committed" d="M934.2,221.5 Q720.1,388.4 657.6,652.5" />
+      <path class="link committed" d="M859.2,376.7 Q692.5,467.2 659.8,654.0" />
+      <path class="link committed" d="M993.5,482.2 Q790.8,502.8 674.3,669.9" />
+      <path class="link committed" d="M897.9,389.0 Q828.5,529.8 901.1,669.0" />
+      <path class="link committed" d="M1055.7,353.1 Q913.4,481.4 920.7,672.9" />
+      <path class="link committed" d="M875.7,385.3 Q743.7,517.6 764.9,703.2" />
+      <path class="link committed" d="M1176.6,486.9 Q1094.0,650.6 1166.6,819.0" />
+      <path class="link committed" d="M1148.6,231.0 Q1066.2,527.3 1163.8,819.0" />
+      <path class="link committed" d="M1293.5,215.2 Q1098.4,405.4 1064.5,675.7" />
+      <path class="link committed" d="M651.2,196.0 Q844.1,511.7 1163.1,699.0" />
+      <path class="link committed" d="M1305.3,450.9 Q1201.1,546.7 1215.5,687.5" />
     </g>
     <g id="nodes">
-      <g class="node full" data-capability="analysis/evidence/calibrate-departures" transform="translate(363.4 689.4)">
+      <g class="node partial" data-capability="analysis/evidence/calibrate-departures" transform="translate(331.4 342.5)">
         <title>Understand how much history supports a finding</title>
         <circle class="node-halo" r="76" />
         <circle class="node-ring" r="68" />
@@ -130,18 +135,18 @@
         </text>
         <text class="meta" y="56">analysis/evidence</text>
       </g>
-      <g class="node full" data-capability="data/inputs/accept-analysis-requests" transform="translate(772.4 181.0)">
+      <g class="node partial" data-capability="data/inputs/accept-analysis-requests" transform="translate(797.8 186.0)">
         <title>Bring my metrics into an analysis</title>
-        <circle class="node-halo" r="71" />
-        <circle class="node-ring" r="63" />
-        <circle class="node-core" r="53" />
+        <circle class="node-halo" r="76" />
+        <circle class="node-ring" r="68" />
+        <circle class="node-core" r="58" />
         <text class="label" y="-3">
           <tspan x="0" dy="-3">Bring my metrics</tspan>
           <tspan x="0" dy="14">into an a...</tspan>
         </text>
-        <text class="meta" y="51">data/inputs</text>
+        <text class="meta" y="56">data/inputs</text>
       </g>
-      <g class="node partial" data-capability="interfaces/cli/analyze-files-and-streams" transform="translate(729.0 363.2)">
+      <g class="node partial" data-capability="interfaces/cli/analyze-files-and-streams" transform="translate(897.1 324.0)">
         <title>Analyze local files from the command line</title>
         <circle class="node-halo" r="66" />
         <circle class="node-ring" r="58" />
@@ -152,7 +157,7 @@
         </text>
         <text class="meta" y="46">interfaces/cli</text>
       </g>
-      <g class="node full" data-capability="analysis/expectations/model-seasonal-level-and-growth" transform="translate(127.0 171.0)">
+      <g class="node partial" data-capability="analysis/expectations/model-seasonal-level-and-growth" transform="translate(421.5 171.0)">
         <title>Account for ordinary seasonal patterns and growth</title>
         <circle class="node-halo" r="61" />
         <circle class="node-ring" r="53" />
@@ -163,7 +168,7 @@
         </text>
         <text class="meta" y="41">analysis/expectations</text>
       </g>
-      <g class="node partial" data-capability="interfaces/python/analyze-through-public-api" transform="translate(958.3 171.0)">
+      <g class="node partial" data-capability="interfaces/python/analyze-through-public-api" transform="translate(966.6 171.0)">
         <title>Embed local analysis in my application</title>
         <circle class="node-halo" r="61" />
         <circle class="node-ring" r="53" />
@@ -174,7 +179,7 @@
         </text>
         <text class="meta" y="41">interfaces/python</text>
       </g>
-      <g class="node full" data-capability="investigation/cases/group-related-evidence" transform="translate(1146.4 171.0)">
+      <g class="node partial" data-capability="investigation/cases/group-related-evidence" transform="translate(1147.0 171.0)">
         <title>Investigate related findings together</title>
         <circle class="node-halo" r="61" />
         <circle class="node-ring" r="53" />
@@ -185,7 +190,18 @@
         </text>
         <text class="meta" y="41">investigation/cases</text>
       </g>
-      <g class="node full" data-capability="data/quality/prepare-timestamped-series" transform="translate(293.7 197.1)">
+      <g class="node partial" data-capability="analysis/detection/identify-residual-patterns" transform="translate(282.7 166.0)">
+        <title>Spot sustained shifts and changing patterns</title>
+        <circle class="node-halo" r="56" />
+        <circle class="node-ring" r="48" />
+        <circle class="node-core" r="38" />
+        <text class="label" y="-3">
+          <tspan x="0" dy="-3">Spot sustained shifts</tspan>
+          <tspan x="0" dy="14">and ...</tspan>
+        </text>
+        <text class="meta" y="36">analysis/detection</text>
+      </g>
+      <g class="node partial" data-capability="data/quality/prepare-timestamped-series" transform="translate(228.9 647.0)">
         <title>Analyze observations in the right time order</title>
         <circle class="node-halo" r="56" />
         <circle class="node-ring" r="48" />
@@ -196,7 +212,7 @@
         </text>
         <text class="meta" y="36">data/quality</text>
       </g>
-      <g class="node partial" data-capability="interfaces/mcp/investigate-findings-and-evidence" transform="translate(1035.1 463.5)">
+      <g class="node partial" data-capability="interfaces/mcp/investigate-findings-and-evidence" transform="translate(1077.1 302.4)">
         <title>Let an agent inspect the evidence behind a finding</title>
         <circle class="node-halo" r="56" />
         <circle class="node-ring" r="48" />
@@ -207,7 +223,7 @@
         </text>
         <text class="meta" y="36">interfaces/mcp</text>
       </g>
-      <g class="node full" data-capability="investigation/assessments/describe-early-changes" transform="translate(1318.0 321.0)">
+      <g class="node partial" data-capability="investigation/assessments/describe-early-changes" transform="translate(1238.9 277.9)">
         <title>Spot early changes with a short history</title>
         <circle class="node-halo" r="56" />
         <circle class="node-ring" r="48" />
@@ -218,7 +234,7 @@
         </text>
         <text class="meta" y="36">investigation/assessments</text>
       </g>
-      <g class="node full" data-capability="relationships/quality/require-compatible-sources" transform="translate(1179.3 397.3)">
+      <g class="node partial" data-capability="relationships/quality/require-compatible-sources" transform="translate(1318.0 166.0)">
         <title>Compare the same entity and time period</title>
         <circle class="node-halo" r="56" />
         <circle class="node-ring" r="48" />
@@ -229,18 +245,29 @@
         </text>
         <text class="meta" y="36">relationships/quality</text>
       </g>
-      <g class="node full" data-capability="analysis/detection/identify-residual-patterns" transform="translate(117.0 305.9)">
-        <title>Spot sustained shifts and changing patterns</title>
+      <g class="node partial" data-capability="analysis/expectations/reset-declared-regimes" transform="translate(117.0 161.0)">
+        <title>Compare against a new baseline after a known change</title>
         <circle class="node-halo" r="51" />
         <circle class="node-ring" r="43" />
         <circle class="node-core" r="33" />
         <text class="label" y="-3">
-          <tspan x="0" dy="-3">Spot sustained shifts</tspan>
-          <tspan x="0" dy="14">and ...</tspan>
+          <tspan x="0" dy="-3">Compare against a</tspan>
+          <tspan x="0" dy="14">new base...</tspan>
         </text>
-        <text class="meta" y="31">analysis/detection</text>
+        <text class="meta" y="31">analysis/expectations</text>
       </g>
-      <g class="node full" data-capability="data/quality/exclude-incomplete-periods" transform="translate(606.5 273.0)">
+      <g class="node partial" data-capability="data/quality/analyze-sparse-observations" transform="translate(615.6 161.0)">
+        <title>Assess sparse observations without inventing consumption</title>
+        <circle class="node-halo" r="51" />
+        <circle class="node-ring" r="43" />
+        <circle class="node-core" r="33" />
+        <text class="label" y="-3">
+          <tspan x="0" dy="-3">Assess sparse</tspan>
+          <tspan x="0" dy="14">observations...</tspan>
+        </text>
+        <text class="meta" y="31">data/quality</text>
+      </g>
+      <g class="node partial" data-capability="data/quality/exclude-incomplete-periods" transform="translate(690.8 314.4)">
         <title>Avoid mistaking unfinished periods for drops</title>
         <circle class="node-halo" r="51" />
         <circle class="node-ring" r="43" />
@@ -251,7 +278,7 @@
         </text>
         <text class="meta" y="31">data/quality</text>
       </g>
-      <g class="node partial" data-capability="interfaces/mcp/connect-read-only-local-server" transform="translate(896.7 331.5)">
+      <g class="node partial" data-capability="interfaces/mcp/connect-read-only-local-server" transform="translate(1036.6 456.9)">
         <title>Give an agent read-only access to local analysis</title>
         <circle class="node-halo" r="51" />
         <circle class="node-ring" r="43" />
@@ -262,7 +289,7 @@
         </text>
         <text class="meta" y="31">interfaces/mcp</text>
       </g>
-      <g class="node full" data-capability="investigation/rules/check-declared-numerical-rules" transform="translate(1057.4 295.9)">
+      <g class="node partial" data-capability="investigation/rules/check-declared-numerical-rules" transform="translate(1178.1 436.9)">
         <title>Find values that break my stated limits</title>
         <circle class="node-halo" r="51" />
         <circle class="node-ring" r="43" />
@@ -273,7 +300,18 @@
         </text>
         <text class="meta" y="31">investigation/rules</text>
       </g>
-      <g class="node full" data-capability="analysis/detection/report-point-anomaly-episodes" transform="translate(112.0 868.0)">
+      <g class="node partial" data-capability="relationships/comparisons/compare-declared-metrics" transform="translate(1323.0 404.2)">
+        <title>Find changes in conversion, unit cost, and metric gaps</title>
+        <circle class="node-halo" r="51" />
+        <circle class="node-ring" r="43" />
+        <circle class="node-core" r="33" />
+        <text class="label" y="-3">
+          <tspan x="0" dy="-3">Find changes</tspan>
+          <tspan x="0" dy="14">in conversion...</tspan>
+        </text>
+        <text class="meta" y="31">relationships/comparisons</text>
+      </g>
+      <g class="node partial" data-capability="analysis/detection/report-point-anomaly-episodes" transform="translate(112.0 868.0)">
         <title>Find unusual spikes and drops</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -284,7 +322,7 @@
         </text>
         <text class="meta" y="26">analysis/detection</text>
       </g>
-      <g class="node full" data-capability="analysis/evidence/assess-reference-readiness" transform="translate(595.3 156.0)">
+      <g class="node partial" data-capability="analysis/evidence/assess-reference-readiness" transform="translate(183.7 268.4)">
         <title>Know when the expected baseline is weak</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -295,7 +333,7 @@
         </text>
         <text class="meta" y="26">analysis/evidence</text>
       </g>
-      <g class="node full" data-capability="analysis/expectations/adapt-to-changing-growth" transform="translate(112.0 449.0)">
+      <g class="node partial" data-capability="analysis/expectations/adapt-to-changing-growth" transform="translate(112.0 572.7)">
         <title>Follow evolving growth while exposing unusual values</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -306,7 +344,7 @@
         </text>
         <text class="meta" y="26">analysis/expectations</text>
       </g>
-      <g class="node uncovered" data-capability="analysis/expectations/model-multiple-seasonal-cycles" transform="translate(242.5 537.0)">
+      <g class="node uncovered" data-capability="analysis/expectations/model-multiple-seasonal-cycles" transform="translate(121.0 403.2)">
         <title>Account for daily and weekly patterns together</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -317,7 +355,7 @@
         </text>
         <text class="meta" y="35">analysis/expectations</text>
       </g>
-      <g class="node full" data-capability="analysis/expectations/protect-reference-from-extremes" transform="translate(112.0 766.6)">
+      <g class="node partial" data-capability="analysis/expectations/protect-reference-from-extremes" transform="translate(535.8 272.3)">
         <title>Limit how extreme values distort later comparisons</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -328,18 +366,7 @@
         </text>
         <text class="meta" y="26">analysis/expectations</text>
       </g>
-      <g class="node full" data-capability="analysis/expectations/reset-declared-regimes" transform="translate(444.5 156.0)">
-        <title>Compare against a new baseline after a known change</title>
-        <circle class="node-halo" r="46" />
-        <circle class="node-ring" r="38" />
-        <circle class="node-core" r="28" />
-        <text class="label" y="-3">
-          <tspan x="0" dy="-3">Compare against a</tspan>
-          <tspan x="0" dy="14">new base...</tspan>
-        </text>
-        <text class="meta" y="26">analysis/expectations</text>
-      </g>
-      <g class="node full" data-capability="analysis/expectations/select-season-length" transform="translate(112.0 587.7)">
+      <g class="node partial" data-capability="analysis/expectations/select-season-length" transform="translate(112.0 724.9)">
         <title>Choose the repeating cycle for a fair comparison</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -350,7 +377,7 @@
         </text>
         <text class="meta" y="26">analysis/expectations</text>
       </g>
-      <g class="node uncovered" data-capability="analysis/rates/assess-bounded-count-rates" transform="translate(267.0 368.4)">
+      <g class="node uncovered" data-capability="analysis/rates/assess-bounded-count-rates" transform="translate(370.3 531.3)">
         <title>Interpret rate changes with their supporting volume</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -361,7 +388,7 @@
         </text>
         <text class="meta" y="35">analysis/rates</text>
       </g>
-      <g class="node full" data-capability="analysis/resolution/monitor-within-unfinished-days" transform="translate(224.6 672.9)">
+      <g class="node partial" data-capability="analysis/resolution/monitor-within-unfinished-days" transform="translate(229.4 499.4)">
         <title>See changes before the day is complete</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -372,7 +399,7 @@
         </text>
         <text class="meta" y="26">analysis/resolution</text>
       </g>
-      <g class="node uncovered" data-capability="data/calendar/aggregate-local-calendar-periods" transform="translate(271.5 859.0)">
+      <g class="node uncovered" data-capability="data/calendar/aggregate-local-calendar-periods" transform="translate(393.9 859.0)">
         <title>Compare days and weeks in my business timezone</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -383,7 +410,7 @@
         </text>
         <text class="meta" y="35">data/calendar</text>
       </g>
-      <g class="node uncovered" data-capability="data/cohorts/analyze-declared-entity-batches" transform="translate(432.8 304.8)">
+      <g class="node uncovered" data-capability="data/cohorts/analyze-declared-entity-batches" transform="translate(512.5 418.6)">
         <title>Analyze many entities without mixing their histories</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -394,7 +421,7 @@
         </text>
         <text class="meta" y="35">data/cohorts</text>
       </g>
-      <g class="node uncovered" data-capability="data/quality/report-source-freshness" transform="translate(452.1 859.0)">
+      <g class="node uncovered" data-capability="data/quality/report-source-freshness" transform="translate(534.9 578.6)">
         <title>Know when missing updates weaken the conclusion</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -405,7 +432,7 @@
         </text>
         <text class="meta" y="35">data/quality</text>
       </g>
-      <g class="node uncovered" data-capability="data/revisions/reconcile-late-and-corrected-observations" transform="translate(391.4 483.7)">
+      <g class="node uncovered" data-capability="data/revisions/reconcile-late-and-corrected-observations" transform="translate(237.1 818.9)">
         <title>See how late or corrected data changes the evidence</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -416,7 +443,7 @@
         </text>
         <text class="meta" y="35">data/revisions</text>
       </g>
-      <g class="node uncovered" data-capability="evaluation/comparisons/compare-monitor-configurations" transform="translate(566.2 415.5)">
+      <g class="node uncovered" data-capability="evaluation/comparisons/compare-monitor-configurations" transform="translate(671.0 476.4)">
         <title>Choose settings using held-out historical evidence</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -427,7 +454,7 @@
         </text>
         <text class="meta" y="35">evaluation/comparisons</text>
       </g>
-      <g class="node uncovered" data-capability="evaluation/outcomes/compare-findings-with-reviewed-incidents" transform="translate(629.7 859.0)">
+      <g class="node uncovered" data-capability="evaluation/outcomes/compare-findings-with-reviewed-incidents" transform="translate(386.9 687.7)">
         <title>Measure detection timing against reviewed incidents</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -438,7 +465,7 @@
         </text>
         <text class="meta" y="35">evaluation/outcomes</text>
       </g>
-      <g class="node uncovered" data-capability="evaluation/replay/replay-historical-observations" transform="translate(526.2 582.7)">
+      <g class="node uncovered" data-capability="evaluation/replay/replay-historical-observations" transform="translate(509.2 780.2)">
         <title>Replay history as it would have arrived</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -449,7 +476,7 @@
         </text>
         <text class="meta" y="35">evaluation/replay</text>
       </g>
-      <g class="node uncovered" data-capability="execution/incremental/append-finalized-observations" transform="translate(862.4 480.6)">
+      <g class="node uncovered" data-capability="execution/incremental/append-finalized-observations" transform="translate(666.0 859.0)">
         <title>Update evidence when the next observations arrive</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -460,7 +487,7 @@
         </text>
         <text class="meta" y="35">execution/incremental</text>
       </g>
-      <g class="node full" data-capability="execution/resilience/retain-evidence-under-failure-and-limits" transform="translate(535.1 747.5)">
+      <g class="node partial" data-capability="execution/resilience/retain-evidence-under-failure-and-limits" transform="translate(888.0 481.8)">
         <title>Keep useful evidence when analysis cannot finish</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -471,7 +498,7 @@
         </text>
         <text class="meta" y="26">execution/resilience</text>
       </g>
-      <g class="node uncovered" data-capability="execution/state/save-and-restore-analysis-state" transform="translate(689.6 539.6)">
+      <g class="node uncovered" data-capability="execution/state/save-and-restore-analysis-state" transform="translate(792.0 591.0)">
         <title>Resume local analysis after a restart</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -482,7 +509,7 @@
         </text>
         <text class="meta" y="35">execution/state</text>
       </g>
-      <g class="node partial" data-capability="interfaces/cli/explain-findings-in-readable-output" transform="translate(900.7 742.7)">
+      <g class="node partial" data-capability="interfaces/cli/explain-findings-in-readable-output" transform="translate(972.4 592.4)">
         <title>Understand findings in a readable summary</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -493,7 +520,7 @@
         </text>
         <text class="meta" y="26">interfaces/cli</text>
       </g>
-      <g class="node uncovered" data-capability="interfaces/export/package-investigation-evidence" transform="translate(1171.0 565.6)">
+      <g class="node uncovered" data-capability="interfaces/export/package-investigation-evidence" transform="translate(1132.5 591.2)">
         <title>Share an investigation that can be inspected later</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -504,7 +531,7 @@
         </text>
         <text class="meta" y="35">interfaces/export</text>
       </g>
-      <g class="node partial" data-capability="interfaces/python/install-versioned-library" transform="translate(680.1 718.5)">
+      <g class="node partial" data-capability="interfaces/python/install-versioned-library" transform="translate(629.5 696.3)">
         <title>Install a versioned analysis library</title>
         <circle class="node-halo" r="49" />
         <circle class="node-ring" r="41" />
@@ -515,7 +542,7 @@
         </text>
         <text class="meta" y="29">interfaces/python</text>
       </g>
-      <g class="node partial" data-capability="interfaces/skill/guide-anomaly-investigation" transform="translate(810.8 633.6)">
+      <g class="node partial" data-capability="interfaces/skill/guide-anomaly-investigation" transform="translate(901.6 718.0)">
         <title>Ask an agent to explain unusual changes in my metrics</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -526,7 +553,7 @@
         </text>
         <text class="meta" y="26">interfaces/skill</text>
       </g>
-      <g class="node partial" data-capability="interfaces/skill/visualize-analysis-on-request" transform="translate(955.0 603.6)">
+      <g class="node partial" data-capability="interfaces/skill/visualize-analysis-on-request" transform="translate(748.8 749.5)">
         <title>See the evidence in a chart when I ask for one</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -537,7 +564,7 @@
         </text>
         <text class="meta" y="26">interfaces/skill</text>
       </g>
-      <g class="node uncovered" data-capability="investigation/cohorts/explain-additive-segment-contributions" transform="translate(1319.0 165.0)">
+      <g class="node uncovered" data-capability="investigation/cohorts/explain-additive-segment-contributions" transform="translate(1319.0 859.0)">
         <title>Find which segments contribute to an aggregate change</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -548,7 +575,7 @@
         </text>
         <text class="meta" y="35">investigation/cohorts</text>
       </g>
-      <g class="node uncovered" data-capability="investigation/history/record-reviewed-case-outcomes" transform="translate(797.9 859.0)">
+      <g class="node uncovered" data-capability="investigation/history/record-reviewed-case-outcomes" transform="translate(1299.6 546.6)">
         <title>Keep reviewed outcomes beside the original evidence</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -559,7 +586,7 @@
         </text>
         <text class="meta" y="35">investigation/history</text>
       </g>
-      <g class="node uncovered" data-capability="investigation/history/retrieve-comparable-reviewed-cases" transform="translate(1319.0 490.3)">
+      <g class="node uncovered" data-capability="investigation/history/retrieve-comparable-reviewed-cases" transform="translate(835.8 859.0)">
         <title>Find earlier reviewed cases with comparable evidence</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -570,7 +597,7 @@
         </text>
         <text class="meta" y="35">investigation/history</text>
       </g>
-      <g class="node full" data-capability="investigation/policy/replay-action-eligibility" transform="translate(1179.5 717.3)">
+      <g class="node partial" data-capability="investigation/policy/replay-action-eligibility" transform="translate(1165.1 868.0)">
         <title>Change action eligibility without rerunning analysis</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -581,7 +608,7 @@
         </text>
         <text class="meta" y="26">investigation/policy</text>
       </g>
-      <g class="node uncovered" data-capability="investigation/regimes/propose-reviewed-baseline-changes" transform="translate(1009.4 859.0)">
+      <g class="node uncovered" data-capability="investigation/regimes/propose-reviewed-baseline-changes" transform="translate(996.5 859.0)">
         <title>Review a possible new baseline before accepting it</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -592,7 +619,7 @@
         </text>
         <text class="meta" y="35">investigation/regimes</text>
       </g>
-      <g class="node uncovered" data-capability="relationships/cohorts/separate-rate-and-population-mix" transform="translate(1180.8 859.0)">
+      <g class="node uncovered" data-capability="relationships/cohorts/separate-rate-and-population-mix" transform="translate(1319.0 723.2)">
         <title>Distinguish changing performance from a changing population mix</title>
         <circle class="node-halo" r="55" />
         <circle class="node-ring" r="47" />
@@ -603,18 +630,7 @@
         </text>
         <text class="meta" y="35">relationships/cohorts</text>
       </g>
-      <g class="node full" data-capability="relationships/comparisons/compare-declared-metrics" transform="translate(1043.1 712.2)">
-        <title>Find changes in conversion, unit cost, and metric gaps</title>
-        <circle class="node-halo" r="46" />
-        <circle class="node-ring" r="38" />
-        <circle class="node-core" r="28" />
-        <text class="label" y="-3">
-          <tspan x="0" dy="-3">Find changes</tspan>
-          <tspan x="0" dy="14">in conversion...</tspan>
-        </text>
-        <text class="meta" y="26">relationships/comparisons</text>
-      </g>
-      <g class="node full" data-capability="relationships/comparisons/evaluate-delayed-response" transform="translate(1328.0 822.1)">
+      <g class="node partial" data-capability="relationships/comparisons/evaluate-delayed-response" transform="translate(1042.7 719.6)">
         <title>Check whether a metric follows its expected delayed response</title>
         <circle class="node-halo" r="46" />
         <circle class="node-ring" r="38" />
@@ -625,16 +641,16 @@
         </text>
         <text class="meta" y="26">relationships/comparisons</text>
       </g>
-      <g class="node uncovered" data-capability="relationships/resolution/compare-finalized-subperiods" transform="translate(1319.0 658.0)">
+      <g class="node partial" data-capability="relationships/resolution/compare-finalized-subperiods" transform="translate(1198.1 733.3)">
         <title>Inspect related metrics before the business period closes</title>
-        <circle class="node-halo" r="55" />
-        <circle class="node-ring" r="47" />
-        <circle class="node-core" r="37" />
+        <circle class="node-halo" r="46" />
+        <circle class="node-ring" r="38" />
+        <circle class="node-core" r="28" />
         <text class="label" y="-3">
           <tspan x="0" dy="-3">Inspect related</tspan>
           <tspan x="0" dy="14">metrics be...</tspan>
         </text>
-        <text class="meta" y="35">relationships/resolution</text>
+        <text class="meta" y="26">relationships/resolution</text>
       </g>
     </g>
   </g>
diff --git a/.capabilities/dependency-viewer.html b/.capabilities/dependency-viewer.html
index 87846aa..144ae27 100644
--- a/.capabilities/dependency-viewer.html
+++ b/.capabilities/dependency-viewer.html
@@ -388,7 +388,7 @@ <h1>Capability dependency viewer</h1>
       <aside><div id="panel" class="panel"></div></aside>
     </div>
     <script>
-const graph = {"nodes":[{"id":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","label":"Understand how much histor...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","acceptance":["Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","Provisional residual scores can be reported but do not trigger point-anomaly episodes.","Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger."],"guidance":["Evidence strength is not a probability of an incident.","Calibration and expectations are separate concerns; each declared regime calibrates again."],"path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/detection/identify-residual-patterns","analysis/detection/report-point-anomaly-episodes","analysis/evidence/assess-reference-readiness","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","relationships/comparisons/compare-declared-metrics"],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"short-history","description":"Keep reference-only evidence for a short series.","command":"python -m pytest -q tests/test_analysis.py::test_short_series_returns_reference_evidence"},{"id":"provisional-evidence","description":"Report provisional scores without point triggers.","command":"python -m pytest -q tests/test_analysis.py::test_early_evidence_maturity_and_no_provisional_trigger"},{"id":"independent-calibration","description":"Check a known residual center, sample scale, and standardized score.","command":"python -m pytest -q tests/test_analysis.py::test_calibration_reference"},{"id":"constant-reference","description":"Handle zero residual variance without spurious episodes.","command":"python -m pytest -q tests/test_analysis.py::test_constant_and_schema"}],"manualChecks":["Inspect a sample's signal_maturity, calibration_samples, standardized_residual, and triggers together before describing a statistical departure."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","criteria":[{"criterion":"Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","status":"covered","evidence":["src/anomalyzer/recipes.py:205","src/anomalyzer/recipes.py:445","tests/test_analysis.py:141"],"notes":"The short-series test checks three reference-only samples, absent standardized scores, and no triggers."},{"criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"covered","evidence":["src/anomalyzer/recipes.py:210","tests/test_analysis.py:123"],"notes":"A provisional score of 2 exceeds the configured point threshold of 1 but produces no trigger or episode."},{"criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"covered","evidence":["src/anomalyzer/recipes.py:231","src/anomalyzer/recipes.py:307","tests/test_analysis.py:111"],"notes":"Calibration uses retained residual mean and sample standard deviation once per regime; the independent numerical example produces the expected standardized score."},{"criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"covered","evidence":["src/anomalyzer/contracts.py:104","src/anomalyzer/recipes.py:350","src/anomalyzer/recipes.py:388","tests/test_analysis.py:111"],"notes":"Sample fields retain the comparison and support; the point threshold is inspectable in the enclosing method parameters."}],"ignore_findings":[],"done":true,"input_fingerprint":"ccdc1b2e92f3e647d5a9e6127cc3c4b10bf0da8f1f78af3efb805daa9b1e4305","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Establish evidence support"},"impact":6,"gaps":0,"r":68,"x":170,"y":170},{"id":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","label":"Bring my metrics into an a...","area":"data","scopes":["data","data/inputs"],"scope":"data/inputs","scopeLabel":"data/inputs","status":"implemented","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","acceptance":["Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","Published schema 1.0 and 1.1 request and result schemas match the public Python contracts."],"guidance":["Null observations represent missing values, never zero. Limits apply to the request as a whole.","Describe current supported behavior; historical plans do not add accepted inputs."],"path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["analysis/evidence/calibrate-departures","analysis/expectations/model-seasonal-level-and-growth","execution/resilience/retain-evidence-under-failure-and-limits","interfaces/python/analyze-through-public-api","investigation/assessments/describe-early-changes"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_agentic.py","schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json"],"automatedChecks":[{"id":"input-forms","description":"Analyze a bare list and validate a canonical request and result.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings tests/test_analysis.py::test_constant_and_schema"},{"id":"explicit-scope","description":"Reject schema 1.1 datasets without entity scope.","command":"python -m pytest -q tests/test_agentic.py::test_schema_11_requires_explicit_entity_scope"},{"id":"invalid-inputs","description":"Reject unsupported settings and nonnumeric or nonfinite observations.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_config tests/test_analysis.py::test_invalid_values"},{"id":"published-schemas","description":"Compare all four checked-in schemas with their public contracts.","command":"python -m pytest -q tests/test_analysis.py::test_checked_in_schemas tests/test_agentic.py::test_v11_checked_in_schemas_exist"}],"manualChecks":["Review the input examples in README.md and confirm they describe the two supported schema versions and the four-dataset/four-relationship limit."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","criteria":[{"criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"covered","evidence":["src/anomalyzer/recipes.py:21","src/anomalyzer/contracts.py:96","tests/test_analysis.py:234"],"notes":"Bare-list settings are validated; schema 1.0 limits datasets to exactly one."},{"criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"covered","evidence":["src/anomalyzer/contracts.py:240","src/anomalyzer/contracts.py:335","tests/test_agentic.py:192"],"notes":"Entity scope is required, relationship kinds are discriminated, and references must name distinct known datasets."},{"criterion":"Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:7","src/anomalyzer/contracts.py:71","tests/test_analysis.py:526","tests/test_analysis.py:531"],"notes":"Assertions reject each listed category; missing null values remain a separate accepted representation."},{"criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"covered","evidence":["schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json","tests/test_analysis.py:547","tests/test_agentic.py:317"],"notes":"Tests and independent inspection compare each complete published schema with its contract-generated schema. Python cross-field validators supplement JSON Schema constraints."}],"ignore_findings":[],"done":true,"input_fingerprint":"3546cd8f436812ff11c9352a855926677b55cc6287e40f6a5041bd613173e5b4","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Supply a numeric series"},"impact":5,"gaps":0,"r":63,"x":353.33333333333337,"y":170},{"id":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","label":"Analyze local files from t...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","acceptance":["The methods command lists supported analysis recipes and residual detectors.","Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status."],"guidance":["CSV callers must select one entity and unit before analysis.","Exit 0 includes completed analysis with anomalies, insufficient history, or inapplicability; exit 2 means invalid input, and exit 1 means execution failure or partial execution."],"path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/python/install-versioned-library","interfaces/skill/guide-anomaly-investigation","interfaces/skill/visualize-analysis-on-request"],"implementationReferences":["src/anomalyzer/cli.py","src/anomalyzer/io.py","src/anomalyzer/__main__.py","tests/test_cli.py"],"automatedChecks":[{"id":"file-stdin","description":"Analyze equivalent file and stdin requests.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence"},{"id":"csv-precedence","description":"Validate selected CSV fields and apply flag-over-file-over-embedded precedence.","command":"python -m pytest -q tests/test_cli.py::test_csv_and_precedence"},{"id":"value-list","description":"Analyze a bare JSON array and expose inferred settings.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"},{"id":"output-and-exits","description":"Protect existing output and distinguish invalid options from failed execution.","command":"python -m pytest -q tests/test_cli.py::test_output_protection tests/test_cli.py::test_methods_and_invalid_options tests/test_cli.py::test_size_limit_and_execution_exit"}],"manualChecks":["Run an example in readable and JSON modes and confirm stdout contains results while diagnostics go to stderr."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","criteria":[{"criterion":"The methods command lists supported analysis recipes and residual detectors.","status":"covered","evidence":["src/anomalyzer/cli.py:426","src/anomalyzer/registry.py:15","tests/test_cli.py:27"],"notes":"Discovery tests check all three recipes and ten residual detectors, with structured stdout and no successful-run diagnostics on stderr."},{"criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:239","tests/test_cli.py:13","tests/test_mcp.py:163"],"notes":"The passing file/stdin regression compares request fingerprints and short-history semantics; cross-interface tests compare complete stable numerical output."},{"criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"covered","evidence":["src/anomalyzer/io.py:33","tests/test_cli.py:51"],"notes":"The passing CSV regression covers timestamped missing cells, positional data, and varying unmapped entity rejection."},{"criterion":"Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","status":"covered","evidence":["src/anomalyzer/cli.py:253","tests/test_cli.py:51","tests/test_cli.py:71"],"notes":"The passing file/flag/embedded precedence regression and bare-values regression establish the merge order and validated configuration."},{"criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"covered","evidence":["src/anomalyzer/io.py:69","src/anomalyzer/cli.py:424","tests/test_cli.py:41","tests/test_cli.py:108"],"notes":"Output protection and exit regressions passed, including explicit overwrite; successful manual examples have empty stderr and incomplete-usage exits 1."}],"ignore_findings":[],"done":true,"input_fingerprint":"86e701ccc0a1aefd407f11b4547784fef497fa1b49ddf6bb69db0d08f757dfdc","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Run from files or stdin"},"impact":4,"gaps":0,"r":58,"x":903.3333333333334,"y":170},{"id":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","label":"Account for ordinary seaso...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","acceptance":["A supplied season length selects earlier matching seasonal observations for the reference.","Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected."],"guidance":["Expectations support scoring; they are not a standalone forecasting API."],"path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/expectations/protect-reference-from-extremes","analysis/expectations/reset-declared-regimes","analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py"],"automatedChecks":[{"id":"seasonal-reference","description":"Check independent lagged-reference values.","command":"python -m pytest -q tests/test_analysis.py::test_forecast_independent_reference"},{"id":"growth-and-departures","description":"Keep clean linear and compound growth ordinary while identifying injected departures.","command":"python -m pytest -q tests/test_trend.py::test_trend_and_anomalies"},{"id":"frozen-fit","description":"Keep the training fit stable when later observations change.","command":"python -m pytest -q tests/test_trend.py::test_frozen_fit_and_legacy_mode"},{"id":"explicit-trends","description":"Accept supported trend choices and handle short, constant, and signed series.","command":"python -m pytest -q tests/test_trend.py::test_explicit_modes_and_validation tests/test_trend.py::test_short_history_constant_and_signed_linear"}],"manualChecks":["Compare examples/growth_up.json and examples/compound.json with their selected trend and evidence expectations."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","criteria":[{"criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"covered","evidence":["src/anomalyzer/baseline.py:8","src/anomalyzer/baseline.py:31","tests/test_analysis.py:21"],"notes":"Independent lag values and robust matching-phase assertions establish the seasonal reference calculation."},{"criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"covered","evidence":["src/anomalyzer/trend.py:120","src/anomalyzer/trend.py:144","tests/test_trend.py:11"],"notes":"Positive and negative linear/exponential cases select the expected model, match post-training expectations, and produce no ordinary-growth episodes."},{"criterion":"The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","status":"covered","evidence":["src/anomalyzer/recipes.py:162","tests/test_trend.py:36"],"notes":"Trend fitting occurs at the training boundary; the test asserts identical trend diagnostics and earlier evidence after later observations change."},{"criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:20","tests/test_trend.py:49","tests/test_trend.py:57"],"notes":"Tests exercise disabled, linear, exponential, and invalid quadratic modes, including disclosed short-history fallback."}],"ignore_findings":[],"done":true,"input_fingerprint":"8e9a0361b02b7d3f651366e5fa0675cccb3949e44de448d2ae0d43e6710bd408","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Account for growth and seasonality"},"impact":3,"gaps":0,"r":53,"x":170,"y":170},{"id":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","label":"Embed local analysis in my...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","acceptance":["analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","analyze_series preserves supplied settings and event-time context for a single dataset.","analyze_relationships preserves the declared multi-dataset relationship semantics.","Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime."],"guidance":["All interfaces use local analysis; there is no hosted service or model API requirement."],"path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence","interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/__init__.py","src/anomalyzer/agent.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"request-parity","description":"Match canonical request results across Python, CLI, and full-result MCP compatibility calls.","command":"python -m pytest -q tests/test_mcp.py::test_canonical_request_matches_python_cli_and_mcp tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"series-context","description":"Preserve a single-series event-time cutoff and context across interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp tests/test_mcp.py::test_series_wrapper_preserves_omitted_and_explicit_season_settings"},{"id":"relationship-parity","description":"Match declared relationship evidence across the public interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_relationship_semantics_match_python_cli_and_mcp"}],"manualChecks":["Review the Python API examples in README.md and confirm callers can inspect the returned validated models and JSON."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","criteria":[{"criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"covered","evidence":["src/anomalyzer/__init__.py:3","src/anomalyzer/recipes.py:21","tests/test_analysis.py:234"],"notes":"Bare values and separate settings validate before dispatch; canonical schema 1.0 and 1.1 requests use their respective contracts."},{"criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"covered","evidence":["src/anomalyzer/agent.py:11","tests/test_mcp.py:191","tests/test_mcp.py:291"],"notes":"Field-set handling preserves omitted versus explicit settings and context in the single-series wrapper, verified across dictionary and Settings inputs."},{"criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"covered","evidence":["src/anomalyzer/agent.py:27","tests/test_mcp.py:217"],"notes":"Declared relationships validate and use the shared engine without changing units, lineage, or primitive semantics."},{"criterion":"Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime.","status":"covered","evidence":["tests/test_mcp.py:163","tests/test_mcp.py:191","tests/test_mcp.py:217","src/anomalyzer/__main__.py:1"],"notes":"Passing real CLI subprocess, Python, and full-result MCP parity tests compare stable numerical results apart from identity and runtime."}],"ignore_findings":[],"done":true,"input_fingerprint":"214bd10bb3aa5fad0f74938fe2ef09853a593cdb5b93cba49ec851a628d8ab65","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","name":"Python integration","label":"Python integration / Embed analysis in my application / Call the public API"},"impact":3,"gaps":0,"r":53,"x":903.3333333333334,"y":170},{"id":"investigation/cases/group-related-evidence","title":"Investigate related findings together","label":"Investigate related findin...","area":"investigation","scopes":["investigation","investigation/cases"],"scope":"investigation/cases","scopeLabel":"investigation/cases","status":"implemented","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","acceptance":["Evidence for different declared entities remains in separate cases even at the same event time.","Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","A derived relationship adds context without counting as an additional independent source for severity.","Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection."],"guidance":["Warning and critical are deterministic numerical grouping rules, not validated incident impact.","Diagnostic trend, oscillation, mixture, and variation findings remain distinct from adaptive location-shift cases."],"path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","dependencies":["investigation/assessments/describe-early-changes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/mcp/investigate-findings-and-evidence","investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/cases.py","src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py"],"automatedChecks":[{"id":"entity-scope","description":"Separate cases for different entities at the same event time.","command":"python -m pytest -q tests/test_agentic.py::test_cases_are_scoped_by_entity_at_the_same_event_time"},{"id":"source-severity","description":"Escalate numerical severity when two source datasets meet criteria.","command":"python -m pytest -q tests/test_agentic.py::test_two_source_datasets_escalate_case_severity"},{"id":"correlated-context","description":"Keep relationship evidence from double-counting source severity.","command":"python -m pytest -q tests/test_agentic.py::test_relationship_evidence_explains_but_does_not_double_count_severity"},{"id":"first-detection","description":"Group an adaptive location shift before point triggers without using a later pattern peak.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_location_shift_creates_one_early_case_without_point_trigger tests/test_agentic.py::test_location_case_severity_does_not_use_future_pattern_peak"}],"manualChecks":["Inspect a case's contributing targets, explanation, evidence references, and investigation questions; distinguish it from a point-anomaly episode."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","criteria":[{"criterion":"Evidence for different declared entities remains in separate cases even at the same event time.","status":"covered","evidence":["src/anomalyzer/cases.py:42","tests/test_agentic.py:69"],"notes":"Entity participates in grouping and stable identity; the test requires separate cases and distinct IDs at the same time."},{"criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"covered","evidence":["src/anomalyzer/cases.py:72","tests/test_agentic.py:85"],"notes":"Two qualifying source datasets produce critical numerical severity and named source explanations."},{"criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"covered","evidence":["src/anomalyzer/cases.py:72","src/anomalyzer/cases.py:116","tests/test_agentic.py:100"],"notes":"The independent-source count includes dataset criteria only. Derived evidence remains explanatory and can still meet the separately declared magnitude rule."},{"criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"covered","evidence":["src/anomalyzer/early.py:193","src/anomalyzer/early.py:237","src/anomalyzer/orchestrator.py:115","tests/test_agentic.py:122","tests/test_agentic.py:146"],"notes":"Tests require one moderate-shift case without point triggers and restrict assessment references/strength to samples at or before first detection."}],"ignore_findings":[],"done":true,"input_fingerprint":"6830af8a2f064123c12adb062f4793e446b97b3b805fd73a934ea84cd0d5d9cf","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Investigate related evidence"},"impact":3,"gaps":0,"r":53,"x":1086.6666666666667,"y":170},{"id":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","label":"Analyze observations in th...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","acceptance":["Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration."],"guidance":["No interpolation, imputation, resampling, or unit conversion occurs.","Duplicate null values propagate through aggregation."],"path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["data/quality/exclude-incomplete-periods","relationships/quality/require-compatible-sources"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/contracts.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"coordinate-contract","description":"Reject partial timestamp metadata and positional as_of cutoffs.","command":"python -m pytest -q tests/test_analysis.py::test_timestamp_metadata_is_all_or_nothing"},{"id":"local-time","description":"Normalize valid local times and reject ambiguous or nonexistent times.","command":"python -m pytest -q tests/test_analysis.py::test_local_time"},{"id":"duplicates-cutoff","description":"Require an explicit duplicate policy and retain counts after the cutoff.","command":"python -m pytest -q tests/test_analysis.py::test_duplicate_sort_and_cutoff"},{"id":"cadence-context","description":"Check fixed weekly cadence and preserve cutoff semantics across interfaces.","command":"python -m pytest -q tests/test_analysis.py::test_weekly_cadence tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp"}],"manualChecks":["Inspect a timestamped result's data_quality transformations, UTC window, and resolved context to confirm preparation is explainable."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","criteria":[{"criterion":"Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","status":"covered","evidence":["src/anomalyzer/contracts.py:77","tests/test_analysis.py:418"],"notes":"Partial coordinate metadata and positional as_of are rejected."},{"criterion":"Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","status":"covered","evidence":["src/anomalyzer/validation.py:8","tests/test_analysis.py:500"],"notes":"UTC round trips detect ambiguous/nonexistent local times; valid local time normalizes to the expected UTC hour."},{"criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"covered","evidence":["src/anomalyzer/validation.py:87","src/anomalyzer/validation.py:103","tests/test_analysis.py:478"],"notes":"Explicit mean/sum handling retains duplicate counts and transformation descriptions. Nulls propagate rather than becoming zero."},{"criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"covered","evidence":["src/anomalyzer/validation.py:105","src/anomalyzer/validation.py:111","tests/test_analysis.py:491","tests/test_mcp.py:191"],"notes":"Filtering is inclusive through as_of; cadence compares UTC elapsed timedeltas, including fixed seven-day weeks."}],"ignore_findings":[],"done":true,"input_fingerprint":"791748764cdc0dfe81d987a908179930d858670606e51b64407172295776b261","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Prepare the comparison"},"impact":2,"gaps":0,"r":48,"x":353.33333333333337,"y":170},{"id":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","label":"Let an agent inspect the e...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","acceptance":["The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","get_evidence exposes the selected finding's numerical support and relationship source lineage.","Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated."],"guidance":["Run results are retained only in the bounded server-session cache; missing or evicted run IDs require reanalysis.","Cases, episodes, and residual patterns may overlap."],"path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":["interfaces/mcp/connect-read-only-local-server","interfaces/skill/guide-anomaly-investigation"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py"],"automatedChecks":[{"id":"case-drilldown","description":"Analyze, page findings, and inspect exact conversion case values and lineage.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case tests/test_mcp.py::test_schema_11_episode_is_paged_and_has_relationship_evidence"},{"id":"unavailable-finding","description":"Expose a quality finding when latest relationship evidence is unavailable.","command":"python -m pytest -q tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"},{"id":"early-candidate","description":"Inspect descriptive candidate evidence without inventing a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case tests/test_mcp.py::test_inapplicable_method_keeps_descriptive_candidate_visible"},{"id":"sample-pagination","description":"Page supporting samples for a residual pattern.","command":"python -m pytest -q tests/test_mcp.py::test_get_evidence_pages_pattern_samples"}],"manualChecks":["Follow analyze, list_findings, and get_evidence for the conversion example and check that the headline is supported by the retrieved values."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","criteria":[{"criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"covered","evidence":["src/anomalyzer/mcp.py:170","src/anomalyzer/mcp_views.py:247","tests/test_mcp.py:54"],"notes":"The conversion manual call returns a relationship headline, preview, counts, readiness, and session-scoped run ID."},{"criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:132","src/anomalyzer/mcp_views.py:260","tests/test_mcp.py:313","tests/test_mcp.py:108"],"notes":"Cases, early candidates, episodes, patterns, and quality findings feed a bounded page; relationship episode regression verifies schema 1.1 pagination."},{"criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:273","src/anomalyzer/mcp_views.py:294","tests/test_mcp.py:54","tests/test_mcp.py:313"],"notes":"Manual drilldown returns observed 8/130 versus expected 0.1 and source indexes orders=14 and qualified_visits=14."},{"criterion":"Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:118","src/anomalyzer/mcp_views.py:169","src/anomalyzer/mcp_views.py:337","tests/test_mcp.py:87","tests/test_mcp.py:137","tests/test_mcp.py:341"],"notes":"Passing tests retain short-history and inapplicable-method candidates plus unavailable latest evidence, while pattern samples are paginated."}],"ignore_findings":[],"done":true,"input_fingerprint":"a09b0003780bcd2f9ccf8e3d8ff58b3147c118760b984757d7027b5810fd9bef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Drill into findings and evidence"},"impact":2,"gaps":0,"r":48,"x":927.3333333333334,"y":288},{"id":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","label":"Spot early changes with a ...","area":"investigation","scopes":["investigation","investigation/assessments"],"scope":"investigation/assessments","scopeLabel":"investigation/assessments","status":"implemented","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","acceptance":["The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","Missing values break the comparison history, and zero previous values do not create relative-change evidence.","Early candidates remain non-triggering and ineligible for action even under a permissive policy."],"guidance":["An early candidate does not establish an anomaly, stable baseline, cause, or incident."],"path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence","investigation/rules/check-declared-numerical-rules"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_mcp.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"prefix-invariant-facts","description":"Describe short history causally without creating cases.","command":"python -m pytest -q tests/test_agentic.py::test_short_history_assessments_are_useful_and_prefix_invariant"},{"id":"comparison-boundaries","description":"Reset comparisons at missing observations and avoid dividing by zero.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"candidate-eligibility","description":"Keep candidates ineligible for action and notifications under permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"},{"id":"agent-visibility","description":"Expose a short-history candidate and its claim limits without a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case"}],"manualChecks":["Inspect a candidate's comparison population, maturity, established claim, and not_established text before describing it."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","criteria":[{"criterion":"The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","status":"covered","evidence":["src/anomalyzer/early.py:73","src/anomalyzer/early.py:96","tests/test_agentic.py:35"],"notes":"Assertions establish observation/change/change/departure_candidate for [10,12,11,20]."},{"criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"covered","evidence":["src/anomalyzer/early.py:50","src/anomalyzer/early.py:141","tests/test_agentic.py:35"],"notes":"Each assessment precedes the prior-history update; tests compare complete serialized assessments for every prefix."},{"criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"covered","evidence":["src/anomalyzer/early.py:53","src/anomalyzer/early.py:58","tests/test_capability_behaviors.py:87"],"notes":"The zero comparison has no relative change; after a missing value, comparison samples reset to zero."},{"criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"covered","evidence":["src/anomalyzer/early.py:105","src/anomalyzer/cases.py:35","tests/test_capability_behaviors.py:134","tests/test_mcp.py:341"],"notes":"Candidates have criterion_met=false and remain action/notification-ineligible after policy replay, without creating cases."}],"ignore_findings":[],"done":true,"input_fingerprint":"debdeeeaea559950f53461a4298cb9c2e3a1681f83a43e8a981ed9b51968a4d0","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect early evidence"},"impact":2,"gaps":0,"r":48,"x":1110.6666666666667,"y":288},{"id":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","label":"Compare the same entity an...","area":"relationships","scopes":["relationships","relationships/quality"],"scope":"relationships/quality","scopeLabel":"relationships/quality","status":"implemented","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","acceptance":["Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","Difference relationships require identical declared source units and matching output units.","An invalid relationship is reported as failed while valid sibling dataset results remain available."],"guidance":["A failed or unavailable relationship does not establish that its metric is normal.","No implicit timestamp matching, interpolation, unit conversion, or zero-filling occurs."],"path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"implementationReferences":["src/anomalyzer/alignment.py","src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"exact-alignment","description":"Reject mismatched timestamps, cadence, entity, duplicates, and positional relationship sources.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relationship_requires_exact_source_alignment"},{"id":"compatible-units","description":"Report a difference unit mismatch without suppressing sibling results.","command":"python -m pytest -q tests/test_agentic.py::test_alignment_unit_and_partial_failure_are_specific"},{"id":"unambiguous-declaration","description":"Reject fields that do not belong to the selected relationship kind.","command":"python -m pytest -q tests/test_agentic.py::test_contract_rejects_ambiguous_relationship_fields"},{"id":"healthy-siblings","description":"Retain a healthy dataset's evidence when another dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Read both target-level statuses and errors before interpreting a partial run; check source dataset errors when a relationship source failed."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","criteria":[{"criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"covered","evidence":["src/anomalyzer/alignment.py:42","tests/test_capability_behaviors.py:30"],"notes":"Parameterized assertions require specific errors for timestamp, cadence, and entity mismatches."},{"criterion":"Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","status":"covered","evidence":["src/anomalyzer/alignment.py:23","src/anomalyzer/alignment.py:44","tests/test_capability_behaviors.py:30"],"notes":"Positional sources fail exact alignment; even an explicit duplicate-mean setting is rejected for relationship source lineage."},{"criterion":"Difference relationships require identical declared source units and matching output units.","status":"covered","evidence":["src/anomalyzer/relationships.py:41","tests/test_agentic.py:242"],"notes":"Source and output unit checks are explicit. Tests cover mismatched source units and successful matching units."},{"criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:171","tests/test_capability_behaviors.py:30","tests/test_capability_behaviors.py:159"],"notes":"Relationship errors remain target-specific, healthy dataset results survive, and the overall mixed run is partial."}],"ignore_findings":[],"done":true,"input_fingerprint":"06321ab5806a8276cc3954df5d6f4dd7f059f67600038854569787ff51b420d6","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Align comparable sources"},"impact":2,"gaps":0,"r":48,"x":1294,"y":288},{"id":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","label":"Spot sustained shifts and ...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","acceptance":["Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","Trend, oscillation, and mixture findings retain their specific pattern meanings.","Moving-range findings identify variation changes separately from location changes."],"guidance":["Pattern findings may overlap and are not independent confirmations.","A residual location shift does not establish a change in the full distribution."],"path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence"],"implementationReferences":["src/anomalyzer/evidence.py","src/anomalyzer/recipes.py","tests/test_analysis.py"],"automatedChecks":[{"id":"location-shift","description":"Find moderate increases and decreases without point anomalies.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_patterns_describe_location_shift_without_point_anomaly"},{"id":"accumulating-change","description":"Detect a known CUSUM accumulation at the expected position.","command":"python -m pytest -q tests/test_analysis.py::test_cusum_detects_accumulated_moderate_location_shift"},{"id":"diagnostic-patterns","description":"Distinguish residual trend, oscillation, and mixture rules.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_diagnostics_have_specific_semantics"},{"id":"variation","description":"Describe a moving-range increase as variation rather than location.","command":"python -m pytest -q tests/test_analysis.py::test_moving_range_detects_short_term_variation_increase"}],"manualChecks":["Compare examples/location_shift_up.json with its point episodes and pattern findings, allowing for overlapping residual rules."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","criteria":[{"criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"covered","evidence":["src/anomalyzer/evidence.py:118","src/anomalyzer/evidence.py:189","tests/test_analysis.py:293"],"notes":"Tests assert no point triggers or episodes and exact upward/downward detection positions for Nelson rules and CUSUM."},{"criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"covered","evidence":["src/anomalyzer/evidence.py:211","src/anomalyzer/evidence.py:230","tests/test_analysis.py:350"],"notes":"The known accumulation detects at index 16 and reports peak CUSUM approximately 6, threshold 5, and increase direction."},{"criterion":"Trend, oscillation, and mixture findings retain their specific pattern meanings.","status":"covered","evidence":["src/anomalyzer/evidence.py:70","src/anomalyzer/evidence.py:122","tests/test_analysis.py:340"],"notes":"Separate predicates and descriptions preserve the three meanings and explicitly limit location/distribution claims."},{"criterion":"Moving-range findings identify variation changes separately from location changes.","status":"covered","evidence":["src/anomalyzer/evidence.py:234","tests/test_analysis.py:360"],"notes":"Adjacent standardized-residual differences produce variation_shift findings; the test checks magnitude, detection index, and interpretation."}],"ignore_findings":[],"done":true,"input_fingerprint":"63716eb67db6b73964b79897d8c81f2891df2586ff213c95ea81c5699b8abbea","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Find sustained residual changes"},"impact":1,"gaps":0,"r":43,"x":194,"y":288},{"id":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","label":"Avoid mistaking unfinished...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","acceptance":["Incomplete observations form a trailing suffix after chronological preparation.","The observed values and timestamps of unfinished periods remain available in data quality output.","Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","A relationship with an unavailable latest source reports that unavailability rather than normal behavior."],"guidance":["Completeness is declared by the caller; the engine does not infer collection completeness."],"path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/alignment.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"unfinished-values","description":"Retain an unfinished value without scoring it.","command":"python -m pytest -q tests/test_analysis.py::test_explicit_incomplete_period_is_reported_and_not_scored"},{"id":"trailing-only","description":"Reject incomplete observations followed by completed observations.","command":"python -m pytest -q tests/test_analysis.py::test_incomplete_periods_must_be_trailing"},{"id":"source-and-agent-views","description":"Exclude a partial source sample from both analysis views and expose unavailable relationship evidence.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"}],"manualChecks":["Inspect examples/agentic/incomplete-usage.json and confirm the latest headline explains missing completeness and does not imply normality."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","criteria":[{"criterion":"Incomplete observations form a trailing suffix after chronological preparation.","status":"covered","evidence":["src/anomalyzer/validation.py:44","tests/test_analysis.py:162"],"notes":"Suffix validation runs after timestamp preparation and rejects a completed observation following an incomplete one."},{"criterion":"The observed values and timestamps of unfinished periods remain available in data quality output.","status":"covered","evidence":["src/anomalyzer/validation.py:50","src/anomalyzer/validation.py:114","tests/test_analysis.py:150"],"notes":"Quality records retain original source index, normalized timestamp, observed value, and incomplete status."},{"criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"covered","evidence":["src/anomalyzer/validation.py:53","src/anomalyzer/recipes.py:564","tests/test_analysis.py:150","tests/test_analysis.py:200"],"notes":"Only finalized prepared values enter either scoring path; assertions verify the last scored sample and both multi-resolution counts."},{"criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"covered","evidence":["src/anomalyzer/alignment.py:35","src/anomalyzer/orchestrator.py:150","tests/test_mcp.py:87"],"notes":"Incomplete aligned sources become unavailable values, and latest relationship state/status plus MCP headline explain the lack of evidence."}],"ignore_findings":[],"done":true,"input_fingerprint":"d82ee816fe89a25232e2589587d31bcadfa81be4188a0e1531f4fbfdf0ebcb5e","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Keep unfinished data visible"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","label":"Give an agent read-only ac...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","acceptance":["Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","Analysis tool calls return structured content together with compact text.","Invalid tool input and unknown session run IDs return structured errors rather than fabricated results."],"guidance":["Compatibility tool names remain callable but are not advertised.","Legacy clients receive isError tool results; modern clients receive JSON-RPC invalid-input errors.","The server provides analysis and evidence inspection; the caller owns monitoring and operational workflows."],"path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","dependencies":["interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":["interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py","pyproject.toml"],"automatedChecks":[{"id":"tool-discovery","description":"Advertise the three read-only tools and their schema contracts.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_lists_read_only_tools"},{"id":"connection-protocols","description":"Support legacy stdio initialization and modern stateless discovery.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_stdio_legacy_handshake_and_tool_listing tests/test_mcp.py::test_mcp_modern_discovery_and_stateless_tool_listing"},{"id":"structured-content","description":"Return a compact analysis response with structured content and run ID.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_tool_call_returns_structured_content"},{"id":"invalid-calls","description":"Reject malformed input and unknown session run IDs in both protocols.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_invalid_tool_input_is_a_tool_error tests/test_mcp.py::test_paged_tool_inputs_enforce_schema_in_both_protocols"}],"manualChecks":["Review discovery's interpretation guidance and confirm the client can start anomalyzer-mcp in the installed Python environment."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","criteria":[{"criterion":"Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","status":"covered","evidence":["src/anomalyzer/mcp.py:63","tests/test_mcp.py:27"],"notes":"Discovery exposes three read-only tools with the same strict argument contracts used at dispatch and advertised output schemas."},{"criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"covered","evidence":["src/anomalyzer/mcp.py:228","tests/test_mcp.py:248","tests/test_mcp.py:268"],"notes":"Legacy initialization and modern stateless discovery tests establish protocol metadata and evidence interpretation guidance."},{"criterion":"Analysis tool calls return structured content together with compact text.","status":"covered","evidence":["src/anomalyzer/mcp.py:162","tests/test_mcp.py:39"],"notes":"The manual conversion call returns a compact headline and structured run ID, counts, readiness, and preview."},{"criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"covered","evidence":["src/anomalyzer/mcp.py:156","src/anomalyzer/mcp.py:257","tests/test_mcp.py:235","tests/test_mcp.py:367"],"notes":"Tests reject invalid cursor/limit types, unknown fields, numeric run IDs, and missing session runs in both protocols without fabricated results."}],"ignore_findings":[],"done":true,"input_fingerprint":"7bd92d238f4f9545372c6c35e78aefea60c585bea0a4a8bb99096ab6cef27f96","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Connect a local server"},"impact":1,"gaps":0,"r":43,"x":927.3333333333334,"y":288},{"id":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","label":"Find values that break my ...","area":"investigation","scopes":["investigation","investigation/rules"],"scope":"investigation/rules","scopeLabel":"investigation/rules","status":"implemented","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","acceptance":["Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","Absolute and relative changes must exceed their declared maximum to violate a rule.","Relative-change checks require a nonzero previous value and do not bridge missing observations.","Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them."],"guidance":["An exact rule crossing does not establish cause, an incident, or business impact."],"path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","dependencies":["investigation/assessments/describe-early-changes"],"suggestedDependencies":[],"dependents":["investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"rule-boundaries","description":"Check inclusive range boundaries and strict absolute/relative change thresholds.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_explicit_rules_respect_boundaries"},{"id":"relative-availability","description":"Avoid relative violations across zero or missing values.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"explicit-not-actionable","description":"Create evidence and a case for an exact rule crossing without enabling action by default.","command":"python -m pytest -q tests/test_agentic.py::test_explicit_rule_is_exact_but_not_actionable_without_policy"},{"id":"readable-rule","description":"Identify the exact unit-cost range rule in CLI output.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the unit-cost example's crossed rule and confirm its text distinguishes the declared bound from statistical detection."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","criteria":[{"criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"covered","evidence":["src/anomalyzer/early.py:22","src/anomalyzer/contracts.py:209","tests/test_capability_behaviors.py:74"],"notes":"Boundary values 10 and 20 are accepted; 9 and 21 violate the declared range."},{"criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"covered","evidence":["src/anomalyzer/early.py:25","tests/test_capability_behaviors.py:74"],"notes":"Parameterized tests distinguish equality to the threshold from strict exceedance for both change rules."},{"criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"covered","evidence":["src/anomalyzer/early.py:29","src/anomalyzer/early.py:53","tests/test_capability_behaviors.py:87"],"notes":"Zero/missing predecessors suppress relative comparisons, and missing values reset comparison history."},{"criterion":"Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them.","status":"covered","evidence":["src/anomalyzer/early.py:34","src/anomalyzer/early.py:121","tests/test_agentic.py:49","tests/test_cli.py:134"],"notes":"Violation records retain rule ID/kind/bounds; default eligibility is false and replay requires explicit permission."}],"ignore_findings":[],"done":true,"input_fingerprint":"60d24e84faee9c1e2ae8f9337d17334b37c89305093404da2e84c51526111d0b","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Apply my stated limits"},"impact":1,"gaps":0,"r":43,"x":1110.6666666666667,"y":288},{"id":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","label":"Find unusual spikes and dr...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","acceptance":["A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","Separated point departures remain separate episodes, and positional episode intervals use integer coordinates."],"guidance":["Detection is a numerical departure from the selected expectation, not a claim about cause or business impact."],"path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/evidence.py","tests/test_analysis.py"],"automatedChecks":[{"id":"point-evidence","description":"Detect an injected spike and retain deterministic earlier numerical evidence.","command":"python -m pytest -q tests/test_analysis.py::test_spike_and_no_lookahead"},{"id":"adjacent-run","description":"Describe adjacent point triggers and their contributing positions.","command":"python -m pytest -q tests/test_analysis.py::test_anomaly_stream_detects_consecutive_run"},{"id":"isolated-events","description":"Keep separated anomalies from becoming a consecutive-run pattern.","command":"python -m pytest -q tests/test_analysis.py::test_isolated_anomalies_do_not_form_second_order_pattern"},{"id":"positional-episodes","description":"Use integer intervals for position-only episode evidence.","command":"python -m pytest -q tests/test_analysis.py::test_positional_episode_uses_integer_interval"}],"manualChecks":["Inspect observations and triggering_samples for examples/spike.json and distinguish episode grouping from pattern findings."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","criteria":[{"criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"covered","evidence":["src/anomalyzer/recipes.py:344","src/anomalyzer/recipes.py:388","tests/test_analysis.py:441"],"notes":"The injected spike retains its observed value, expectation, standardized residual, and point trigger."},{"criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"covered","evidence":["src/anomalyzer/baseline.py:31","src/anomalyzer/recipes.py:174","tests/test_analysis.py:441","tests/test_trend.py:36"],"notes":"Assertions establish repeatability and unchanged earlier evidence under later evaluation changes with fixed seasonal settings. Automatic positional season discovery has the separately documented streaming limitation."},{"criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"covered","evidence":["src/anomalyzer/evidence.py:95","tests/test_analysis.py:274"],"notes":"The test checks one three-sample run, exact contributing positions, direction, and first detection."},{"criterion":"Separated point departures remain separate episodes, and positional episode intervals use integer coordinates.","status":"covered","evidence":["src/anomalyzer/evidence.py:12","src/anomalyzer/evidence.py:23","tests/test_analysis.py:263","tests/test_analysis.py:369"],"notes":"Separated spikes produce two episodes without a consecutive-run pattern; positional evidence produces interval [9, 9]."}],"ignore_findings":[],"done":true,"input_fingerprint":"062833cc1bd19d5734b0f3c63d1e81666ffb568ead3399e359d62db2c0c0677d","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Find spikes and drops"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","label":"Know when the expected bas...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","acceptance":["Repeating seasonal structure with stable calibration can yield supported readiness.","Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","Insufficient calibration support yields not_assessed, including when the run otherwise completes.","Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment."],"guidance":["Readiness does not prove real-world effectiveness and is not an automatic action-policy gate."],"path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/readiness.py","src/anomalyzer/recipes.py","tests/test_readiness.py"],"automatedChecks":[{"id":"supported-reference","description":"Support repeated seasonality and preserve readiness under later changes.","command":"python -m pytest -q tests/test_readiness.py::test_repeating_season_and_trend_support_reference"},{"id":"weak-season","description":"Warn on unstable seasonality even when no findings occur.","command":"python -m pytest -q tests/test_readiness.py::test_unstable_declared_season_warns_without_calling_series_inapplicable tests/test_readiness.py::test_documented_random_example_has_no_findings_but_warns_on_reference"},{"id":"calibration-drift","description":"Warn when residual calibration drifts despite a repeated training season.","command":"python -m pytest -q tests/test_readiness.py::test_calibration_drift_warns_even_when_training_season_repeats"},{"id":"not-assessed","description":"Distinguish uncalibrated history and too few calibration samples.","command":"python -m pytest -q tests/test_readiness.py::test_readiness_is_not_assessed_without_calibrated_evaluation tests/test_readiness.py::test_short_calibration_is_not_assessed_even_when_analysis_completes"}],"manualChecks":["Read readiness reasons alongside findings; a lack of findings alone does not establish a trustworthy reference."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","criteria":[{"criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"covered","evidence":["src/anomalyzer/readiness.py:128","tests/test_readiness.py:8"],"notes":"Assertions establish repeated seasonality, strong calibration improvement, and supported readiness."},{"criterion":"Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","status":"covered","evidence":["src/anomalyzer/readiness.py:113","src/anomalyzer/readiness.py:122","tests/test_readiness.py:28","tests/test_readiness.py:41"],"notes":"Completed analyses retain caution reasons for weak seasonal repetition or calibration-center drift."},{"criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"covered","evidence":["src/anomalyzer/readiness.py:77","tests/test_readiness.py:56","tests/test_readiness.py:63"],"notes":"Both uncalibrated history and a completed run with three calibration samples return not_assessed."},{"criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"covered","evidence":["src/anomalyzer/recipes.py:327","tests/test_readiness.py:8"],"notes":"Readiness reads training/calibration inputs once; tests compare identical segment and aggregate assessments after a later shift."}],"ignore_findings":[],"done":true,"input_fingerprint":"d01367627e058ea39c78ee0e3616061a2e0725d06e8acd0b84b46a9655431e57","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Assess baseline quality"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","label":"Follow evolving growth whi...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","acceptance":["Each rolling expectation uses only observations before the sample being scored.","Persistent growth changes can be followed more closely than by the frozen default recipe.","An isolated extreme value is held out of future model references rather than creating a seasonal echo.","Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected."],"guidance":["A persistent third extreme miss in one direction may be admitted to the rolling reference; this is not a manual regime reset."],"path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"causal-window","description":"Keep a known rolling expectation invariant when future values are appended.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_expectation_is_one_step_and_prefix_invariant"},{"id":"changing-growth","description":"Follow a persistent growth change and report model admissions.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_tracks_a_persistent_growth_change"},{"id":"isolated-extreme","description":"Exclude an isolated extreme from later reference estimates.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_contains_an_isolated_extreme"},{"id":"bounded-settings","description":"Reject conflicting or unbounded adaptive settings.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_settings_reject_conflicting_or_unbounded_trend"}],"manualChecks":["Inspect the adaptive recipe's rolling-window parameters and frozen calibration before interpreting its final expectation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","criteria":[{"criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"covered","evidence":["src/anomalyzer/baseline.py:61","tests/test_analysis.py:48"],"notes":"All rolling-reference positions precede the scored index. The independent expectation remains unchanged after future values are appended."},{"criterion":"Persistent growth changes can be followed more closely than by the frozen default recipe.","status":"covered","evidence":["src/anomalyzer/recipes.py:145","src/anomalyzer/recipes.py:370","tests/test_analysis.py:58"],"notes":"The test requires adaptive held-out MAE below half the frozen recipe's MAE and a close final expectation."},{"criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","tests/test_analysis.py:81"],"notes":"The isolated spike is excluded from model references; its next seasonal expectation closely matches the control. Persistent third misses are admitted as guidance explicitly permits."},{"criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:41","tests/test_analysis.py:104"],"notes":"Both incompatible explicit trend selection and excessive slope lookback raise specific validation errors."}],"ignore_findings":[],"done":true,"input_fingerprint":"984422f1fd3ab3d22cf7b4ba9d0bd06629fc228a8e53d7ad44439cd8d35ec0db","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Follow evolving growth"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","label":"Account for daily and week...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"planned","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","acceptance":["Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","Insufficient support for a requested cycle produces an explicit limitation or declared fallback rather than a silently fitted component.","Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle."],"guidance":["Begin with a bounded daily-and-weekly use case; overlapping cycles share information and cannot be treated as independent evidence.","Proposed automated checks cover combined ordinary cycles, future isolation, unsupported cycle history, and extreme-value influence across cycles.","Manual review should inspect a weekday morning spike and verify the baseline explains ordinary calendar variation without explaining away the spike."],"path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Represent more than one repeating cycle"},"impact":0,"gaps":3,"r":47,"x":170,"y":406},{"id":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","label":"Limit how extreme values d...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","acceptance":["Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","Training and calibration diagnostics identify excluded extreme observations.","An extreme evaluated sample remains evidence while its expected value is used in future references.","Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior."],"guidance":["Protection keeps the observed evidence; it does not delete observations or certify that a departure was an incident."],"path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/robust.py","src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py","tests/test_cli.py"],"automatedChecks":[{"id":"robust-reference","description":"Limit an extreme weekly reference and use earlier references at lag one.","command":"python -m pytest -q tests/test_analysis.py::test_robust_weekly_reference_limits_one_high_week tests/test_analysis.py::test_robust_lag_one_uses_prior_seasonal_values"},{"id":"training-calibration","description":"Record training and calibration exclusions and retain a useful growth fit.","command":"python -m pytest -q tests/test_trend.py::test_robust_training_fit_and_calibration_exclusions"},{"id":"include-choice","description":"Compare robust and include modes through the CLI and show their differing triggers.","command":"python -m pytest -q tests/test_cli.py::test_outlier_handling_cli"}],"manualChecks":["Inspect excluded_from_model, reference_action, and outlier-handling diagnostics alongside the original observed value."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","criteria":[{"criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"covered","evidence":["src/anomalyzer/baseline.py:39","tests/test_analysis.py:25"],"notes":"Assertions check a independently computed weighted reference and invariance when an already-trimmed high value becomes more extreme."},{"criterion":"Training and calibration diagnostics identify excluded extreme observations.","status":"covered","evidence":["src/anomalyzer/trend.py:178","src/anomalyzer/recipes.py:287","tests/test_trend.py:65"],"notes":"Tests identify exact training/calibration exclusions and compare their influence on trend selection and residual scale."},{"criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","src/anomalyzer/recipes.py:392","tests/test_analysis.py:263","tests/test_trend.py:11"],"notes":"The observed departure and trigger survive; only reference_values receives the prior expectation, preventing the next seasonal echo."},{"criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"covered","evidence":["src/anomalyzer/recipes.py:151","src/anomalyzer/recipes.py:475","tests/test_cli.py:95"],"notes":"CLI assertions distinguish robust triggers {65} from include-mode triggers {65, 72}."}],"ignore_findings":[],"done":true,"input_fingerprint":"183b8696fdc9786878b3e54c4a9e5f68180b47e157528bef953274a3487d49a7","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Protect later comparisons"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","label":"Compare against a new base...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","acceptance":["Integer positions and exact observed dates or timestamps can identify reset boundaries.","Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","Repeated, out-of-range, or unmatched reset points are rejected.","CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries."],"guidance":["Reset points are explicit caller decisions. Known-event notes and review-state fields do not rebuild regimes.","Multi-resolution analysis does not support reset_points."],"path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py"],"automatedChecks":[{"id":"independent-regimes","description":"Retrain separate regimes without cross-boundary evidence or patterns.","command":"python -m pytest -q tests/test_analysis.py::test_manual_resets_retrain_without_crossing_boundaries"},{"id":"invalid-boundaries","description":"Reject invalid or duplicate resolved reset positions.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_manual_resets"},{"id":"cli-boundaries","description":"Resolve position and date reset flags through the CLI.","command":"python -m pytest -q tests/test_cli.py::test_manual_reset_cli"}],"manualChecks":["Inspect the resolved reset positions and segment diagnostics for examples/steps_reset.json."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","criteria":[{"criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"covered","evidence":["src/anomalyzer/validation.py:129","tests/test_analysis.py:381"],"notes":"Positions and observed dates are exercised end to end; timestamp normalization and exact matching are implemented directly."},{"criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"covered","evidence":["src/anomalyzer/recipes.py:119","src/anomalyzer/recipes.py:143","tests/test_analysis.py:381"],"notes":"Each segment initializes separate state and passes its own start to forecasts. Tests verify three calibrated regimes and no cross-boundary evidence or findings."},{"criterion":"Repeated, out-of-range, or unmatched reset points are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/validation.py:149","tests/test_analysis.py:413"],"notes":"Validation rejects literal duplicates and different declarations resolving to the same position, invalid positions, and unmatched dates/timestamps."},{"criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"covered","evidence":["tests/test_cli.py:82","src/anomalyzer/recipes.py:53"],"notes":"CLI assertions establish positions [50, 100] and removal of the unmarked-regime episodes."}],"ignore_findings":[],"done":true,"input_fingerprint":"067063f972c5b70d990d6cb0125892731a97a7f482698d7b331a1a30229ac223","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Restart after a reviewed change"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","label":"Choose the repeating cycle...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","acceptance":["An explicit season length is honored and reported as explicit.","A repeating positional series can infer a seasonal lag and report the selection window.","A positional series without a sufficiently stable candidate falls back to lag one and records why.","The reserved calibration and evaluation tail does not determine positional season selection."],"guidance":["Candidate discovery uses the input prefix before the reserved tail and confirms the candidate in an early window. It is not fully prefix-invariant streaming discovery.","Supply season_length explicitly when stable streaming season selection is required."],"path":".capabilities/analysis/expectations/select-season-length.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/seasonality.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"declared-season","description":"Honor a separate explicit season setting.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"inference-fallback","description":"Infer a repeating lag or record lag-one fallback when no season is supported.","command":"python -m pytest -q tests/test_analysis.py::test_positional_season_inference_and_fallback"},{"id":"reserved-tail","description":"Keep selection unchanged when only the reserved tail changes.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_season_inference_excludes_reserved_tail"},{"id":"cli-inference","description":"Expose the inferred lag and its source for bare JSON values in the CLI.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"}],"manualChecks":["Review selection_window and ranked candidates for a positional series before treating its inferred season as an ongoing monitoring assumption."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","criteria":[{"criterion":"An explicit season length is honored and reported as explicit.","status":"covered","evidence":["src/anomalyzer/recipes.py:55","tests/test_analysis.py:234"],"notes":"Explicit field presence bypasses inference, and the test checks the reported source."},{"criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"covered","evidence":["src/anomalyzer/seasonality.py:96","tests/test_analysis.py:247"],"notes":"The repeating seven-position series selects lag 7 and reports confirmation window [0, 27]."},{"criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"covered","evidence":["src/anomalyzer/seasonality.py:79","src/anomalyzer/seasonality.py:84","tests/test_analysis.py:247"],"notes":"Candidate-strength and early-confirmation failures return lag 1 with a reason; the quadratic example verifies fallback."},{"criterion":"The reserved calibration and evaluation tail does not determine positional season selection.","status":"covered","evidence":["src/anomalyzer/seasonality.py:69","tests/test_capability_behaviors.py:147"],"notes":"Discovery excludes calibration_size plus one reserved sample. Changing that tail preserves the complete inference diagnostics; arbitrary appended-prefix invariance is explicitly outside the guidance."}],"ignore_findings":[],"done":true,"input_fingerprint":"3317b1727dc3cabfcfc9b87808d9249d0d59f111b8fd5f8a1fad653b21343980","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Choose a repeating cycle"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","label":"Interpret rate changes wit...","area":"analysis","scopes":["analysis","analysis/rates"],"scope":"analysis/rates","scopeLabel":"analysis/rates","status":"planned","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","acceptance":["Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","Zero denominators remain unavailable, and low support is explicitly marked; intervals remain within the valid zero-to-one rate domain.","Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations."],"guidance":["Independence and stable exposure are modeling assumptions; clustered events, overdispersion, or changing populations require explicit limitations.","Proposed automated checks cover zero denominators, tiny counts, zero-or-all-success boundaries, and equal rates at different volumes.","Manual review should compare sparse and high-volume conversion changes and confirm the explanation discloses assumptions without inventing incident probability."],"path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Judge rates using their exposure"},"impact":0,"gaps":3,"r":47,"x":194,"y":524},{"id":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","label":"See changes before the day...","area":"analysis","scopes":["analysis","analysis/resolution"],"scope":"analysis/resolution","scopeLabel":"analysis/resolution","status":"implemented","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","acceptance":["Callers can select sum, mean, or last aggregation for completed and partial periods.","Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","An explicitly unfinished source sample is excluded from both analysis views.","A source cadence that does not evenly divide 24 hours is rejected."],"guidance":["Periods are fixed 24-hour windows, not daylight-saving-aware local calendar days.","Native and daily views share source observations and are correlated.","This recipe supports schema 1.0 and does not support reset_points."],"path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","dependencies":["data/quality/exclude-incomplete-periods","analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"selected-aggregation","description":"Check sum, mean, and last for completed days and trailing partial aggregates.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_multi_resolution_uses_selected_aggregation"},{"id":"two-views","description":"Return completed days, native evidence, and a linked trailing partial-day marker.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_current_period_and_keeps_two_views"},{"id":"unfinished-source","description":"Exclude a marked incomplete source sample from both views.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days"},{"id":"dividing-cadence","description":"Reject subperiod durations that do not evenly divide a day.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_rejects_non_dividing_frequency"}],"manualChecks":["Inspect the partial aggregate's observed_aggregate and intraday_evidence_refs together with the two method namespaces."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","criteria":[{"criterion":"Callers can select sum, mean, or last aggregation for completed and partial periods.","status":"covered","evidence":["src/anomalyzer/contracts.py:18","src/anomalyzer/recipes.py:539","tests/test_capability_behaviors.py:178"],"notes":"Parameterized assertions check independently known completed and partial values for all three aggregation choices."},{"criterion":"Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","status":"covered","evidence":["src/anomalyzer/recipes.py:617","src/anomalyzer/recipes.py:667","src/anomalyzer/recipes.py:680","tests/test_analysis.py:170"],"notes":"Separate datasets, configurations, and namespaces preserve the two correlated views."},{"criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"covered","evidence":["src/anomalyzer/recipes.py:639","src/anomalyzer/recipes.py:690","tests/test_analysis.py:170"],"notes":"The marker contains finalized progress, observed aggregate, native evidence references, and an explicit exclusion flag."},{"criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"covered","evidence":["src/anomalyzer/validation.py:44","src/anomalyzer/recipes.py:564","tests/test_analysis.py:200"],"notes":"Both child datasets use prepared finalized observations; tests verify counts and retention of the excluded original value."},{"criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"covered","evidence":["src/anomalyzer/recipes.py:572","tests/test_analysis.py:221"],"notes":"A five-hour cadence raises the specific dividing-cadence error."}],"ignore_findings":[],"done":true,"input_fingerprint":"cdcdea1166a4e3d2523b594bec938142135e17a6b4279732e52819a486d1b4f2","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Compare finalized subperiods"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","label":"Compare days and weeks in ...","area":"data","scopes":["data","data/calendar"],"scope":"data/calendar","scopeLabel":"data/calendar","status":"planned","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","acceptance":["Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","Periods report UTC boundaries and local labels; daylight-saving days can contain 23 or 25 hours without being forced into fixed 24-hour windows.","Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation."],"guidance":["Calendar grouping needs a distinct contract from fixed elapsed cadence; it must not silently reinterpret existing frequency or aggregation settings.","Proposed automated checks cover spring transitions, fall transitions, week boundaries, and missing native slots in a calendar period.","Manual review should inspect a daylight-saving week and confirm local labels, observed coverage, and completed-versus-open comparisons are understandable."],"path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Declare local calendar boundaries"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","label":"Analyze many entities with...","area":"data","scopes":["data","data/cohorts"],"scope":"data/cohorts","scopeLabel":"data/cohorts","status":"planned","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","acceptance":["Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","Training, calibration, relationships, and cases do not cross entity boundaries unless a separate explicit aggregate comparison is requested.","The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident."],"guidance":["Batch orchestration should preserve existing per-request contracts rather than silently lifting schema limits; multiple-testing limits stay visible.","Proposed automated checks cover partition isolation, batch-versus-single equivalence, one-entity failure, and shared resource-budget exhaustion.","Manual review should inspect a mixed batch of healthy, changed, and sparse entities and confirm the summary does not equate missing support with health."],"path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Keep each entity's evidence separate"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","label":"Know when missing updates ...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"planned","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","acceptance":["Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","Results distinguish current, late, missing, and not-assessed freshness and expose the last observed update and the deadline used.","A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision."],"guidance":["Freshness depends on declared collection expectations; it does not prove source completeness and does not start a scheduler or send a notification.","Proposed automated checks cover exact deadlines, allowed lateness, absent source updates, and stale-source propagation to relationships.","Manual review should compare a stale feed with a true metric drop and confirm the report distinguishes collection uncertainty from numerical departure."],"path":".capabilities/data/quality/report-source-freshness.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Check whether the sources are current"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","label":"See how late or corrected ...","area":"data","scopes":["data","data/revisions"],"scope":"data/revisions","scopeLabel":"data/revisions","status":"planned","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","acceptance":["A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","Results preserve prior and new evidence versions and identify findings that were added, changed, or withdrawn with traceable source lineage.","Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results."],"guidance":["A corrected event-time result does not replace the record of what was known earlier; callers control retention and any notification about a revision.","Proposed automated checks cover late completion, corrected calibration data, downstream relationship changes, and conflicting source versions.","Manual review should inspect a drop withdrawn after ingestion completes and confirm both the original limitation and the revision explanation are retained."],"path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Reconcile revisions without losing history"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","label":"Choose settings using held...","area":"evaluation","scopes":["evaluation","evaluation/comparisons"],"scope":"evaluation/comparisons","scopeLabel":"evaluation/comparisons","status":"planned","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","acceptance":["Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","The output preserves every candidate's settings and evaluation provenance; adopting a configuration requires an explicit caller decision."],"guidance":["Bounded comparison is not unrestricted parameter search; held-out results describe the supplied history and do not guarantee future effectiveness.","Proposed automated checks cover split isolation, comparable replay inputs, failed candidate retention, and unchanged active settings after comparison.","Manual review should compare an earlier but noisier configuration with a quieter one and confirm the tradeoff is visible without an unsupported winner."],"path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Review the tradeoff before adopting settings"},"impact":0,"gaps":3,"r":47,"x":560.6666666666667,"y":524},{"id":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","label":"Measure detection timing a...","area":"evaluation","scopes":["evaluation","evaluation/outcomes"],"scope":"evaluation/outcomes","scopeLabel":"evaluation/outcomes","status":"planned","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","acceptance":["Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","Timing reports distinguish the incident's supplied start from the first available candidate, calibrated detection, and policy-eligible evidence.","Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown."],"guidance":["Detection timing does not measure response effectiveness or establish that an unmatched finding was false; incident labels are caller judgments.","Proposed automated checks cover boundary matching, ambiguous overlaps, incomplete label coverage, and separate candidate versus eligibility timing.","Manual review should inspect a matched and an unmatched episode and confirm the reported timing and accuracy limits agree with the supplied label coverage."],"path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Compare detection with reviewed outcomes"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","label":"Replay history as it would...","area":"evaluation","scopes":["evaluation","evaluation/replay"],"scope":"evaluation/replay","scopeLabel":"evaluation/replay","status":"planned","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","acceptance":["Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","Repeating a replay preserves numerical evidence and first-detection timing apart from generated IDs and runtime; stopped replays retain completed steps."],"guidance":["Event-time replay does not reconstruct arrival-time availability unless the caller supplies versioned arrival history.","Proposed automated checks cover future-tail isolation, short-history cutoffs, deterministic repeated replay, and interrupted replay retention.","Manual review should inspect one change from its first available observation through calibration and confirm the explanation uses only then-known evidence."],"path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Reconstruct the evidence available at the time"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","label":"Update evidence when the n...","area":"execution","scopes":["execution","execution/incremental"],"scope":"execution/incremental","scopeLabel":"execution/incremental","status":"planned","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","acceptance":["Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","Supported explicitly configured recipes produce the same numerical evidence as full batch analysis of the combined history, including cross-chunk patterns.","Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations."],"guidance":["Start with explicit season lengths; input-prefix season discovery is not currently guaranteed to be invariant when the history grows.","Proposed automated checks cover chunk-versus-batch equivalence, cross-chunk CUSUM, idempotent retry, and failed-append state preservation.","Manual review should follow a sustained change across several append batches and confirm detection timing and provisional evidence remain understandable."],"path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Analyze the next finalized samples"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","label":"Keep useful evidence when ...","area":"execution","scopes":["execution","execution/resilience"],"scope":"execution/resilience","scopeLabel":"execution/resilience","status":"implemented","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","acceptance":["A method failure preserves evidence completed before the failure and reports the error.","Runtime budgets are checked cooperatively and produce an explicit stop reason.","CLI byte limits and execution failures produce distinct error exits.","A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial."],"guidance":["Default limits are 10,000 source observations per request, 10 MB CLI input, and 60 seconds cooperative runtime.","Numerical results are deterministic for fixed declared settings; generated run IDs and runtimes vary."],"path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/budget.py","src/anomalyzer/recipes.py","src/anomalyzer/orchestrator.py","src/anomalyzer/io.py","src/anomalyzer/cli.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"method-and-budget","description":"Retain pre-failure evidence and report a runtime-budget stop.","command":"python -m pytest -q tests/test_analysis.py::test_failure_and_budget"},{"id":"cli-limits","description":"Enforce byte limits and distinguish execution failure in CLI output and exit status.","command":"python -m pytest -q tests/test_cli.py::test_size_limit_and_execution_exit"},{"id":"partial-siblings","description":"Preserve healthy analysis when a sibling dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Inspect run status, stop_reason, and each target's status before relying on a result; confirm limits are presented as cooperative rather than hard process timeouts."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","criteria":[{"criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"covered","evidence":["src/anomalyzer/recipes.py:400","src/anomalyzer/recipes.py:448","tests/test_analysis.py:509"],"notes":"The failure regression preserves evidence through index 44 and reports the injected error at index 45."},{"criterion":"Runtime budgets are checked cooperatively and produce an explicit stop reason.","status":"covered","evidence":["src/anomalyzer/budget.py:18","src/anomalyzer/recipes.py:108","src/anomalyzer/recipes.py:508","tests/test_analysis.py:509"],"notes":"Cooperative target/sample checks expose runtime_budget and budget_exceeded; this is not a hard process timeout."},{"criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:429","tests/test_cli.py:108"],"notes":"The full suite executed the byte-limit and execution-exit regression successfully; invalid input exits 2 and execution failure exits 1."},{"criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:125","src/anomalyzer/orchestrator.py:171","src/anomalyzer/orchestrator.py:185","tests/test_capability_behaviors.py:159","tests/test_capability_behaviors.py:30"],"notes":"Healthy dataset evidence survives target-specific failure; inspected incomplete-usage reports partial with no invented healthy relationship."}],"ignore_findings":[],"done":true,"input_fingerprint":"146a9c8396fd7a2d039f75077dc0b82a0778685d20c3782aa284d5f0b6ed94be","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect incomplete execution"},"impact":0,"gaps":0,"r":38,"x":720,"y":642},{"id":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","label":"Resume local analysis afte...","area":"execution","scopes":["execution","execution/state"],"scope":"execution/state","scopeLabel":"execution/state","status":"planned","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","acceptance":["A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","Restoring a supported state and continuing yields the same numerical evidence and detection positions as uninterrupted processing of the same observations.","Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled."],"guidance":["Stored observations and context can contain sensitive data; document exactly what a snapshot contains and avoid automatic persistence.","Proposed automated checks cover restart equivalence, corrupt snapshots, version or settings mismatch, and output overwrite protection.","Manual review should inspect a snapshot and resumed explanation and confirm their provenance and retained-data scope are clear to an application owner."],"path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Preserve state across restarts"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","label":"Understand findings in a r...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","acceptance":["Point-anomaly episode counts and residual-pattern findings are labeled separately.","Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","Readiness status remains visible, including when it was not assessed."],"guidance":["Readable summaries should retain the evidence's claim limits and allow drilldown into JSON."],"path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/cli.py","tests/test_cli.py","examples/agentic/conversion.json","examples/agentic/unit-cost.json"],"automatedChecks":[{"id":"episodes-patterns","description":"Distinguish shift patterns from point episodes in readable output.","command":"python -m pytest -q tests/test_cli.py::test_readable_output_distinguishes_episodes_from_patterns"},{"id":"relationship-case","description":"Explain a conversion case with entity, units, comparison, and a suggested check.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_relationship_request_through_cli"},{"id":"explicit-rule","description":"Describe the unit-cost rule boundary and unassessed readiness.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the conversion and unit-cost example output and confirm the descriptions state the supported numerical comparison in plain language."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","criteria":[{"criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"covered","evidence":["src/anomalyzer/cli.py:347","src/anomalyzer/cli.py:405","tests/test_cli.py:143"],"notes":"Readable tests distinguish zero-episode residual shifts from a spike episode and retain the overlap warning."},{"criterion":"Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","status":"covered","evidence":["src/anomalyzer/cli.py:291","src/anomalyzer/cli.py:308","src/anomalyzer/cli.py:331","tests/test_cli.py:117","examples/agentic/conversion.json"],"notes":"Conversion output identifies site=demo, observed/expected orders_per_visit, the relationship, and a collection-completeness investigation question."},{"criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"covered","evidence":["src/anomalyzer/cli.py:318","tests/test_cli.py:134","examples/agentic/unit-cost.json"],"notes":"Unit-cost output names review-above-13c and its [0, 0.13] boundary as a declared rule crossing."},{"criterion":"Readiness status remains visible, including when it was not assessed.","status":"covered","evidence":["src/anomalyzer/cli.py:352","src/anomalyzer/cli.py:399","tests/test_cli.py:134"],"notes":"Readable output includes not_assessed and its supporting reasons without claiming measured alert accuracy."}],"ignore_findings":[],"done":true,"input_fingerprint":"98108868b000972fd9b3c3f082846fbd3f3b08c399ef7ed424c23245d6935fef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Read the findings"},"impact":0,"gaps":0,"r":38,"x":903.3333333333334,"y":642},{"id":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","label":"Share an investigation tha...","area":"interfaces","scopes":["interfaces","interfaces/export"],"scope":"interfaces/export","scopeLabel":"interfaces/export","status":"planned","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","acceptance":["Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","Source observations and context are included only by explicit selection; omitted data and resulting limits on independent reproduction are disclosed."],"guidance":["Fingerprints identify included provenance; they do not prove source truth. A portable report must make privacy scope and reproduction limits inspectable.","Proposed automated checks cover manifest consistency, partial-run export, omitted-source disclosure, and overwrite protection.","Manual review should open a bundle outside the original session and confirm a reviewer can trace one finding and see exactly what cannot be reproduced."],"path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Preserve a portable evidence record"},"impact":0,"gaps":3,"r":47,"x":903.3333333333334,"y":642},{"id":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","label":"Install a versioned analys...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","acceptance":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"guidance":["Guard owns scheduling, persistence, business severity, incidents and notifications.","A published package does not establish integration into UsageTap Guard."],"path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","interfaces/cli/analyze-files-and-streams","interfaces/mcp/connect-read-only-local-server"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["pyproject.toml","src/anomalyzer/_version.py","scripts/release.py","scripts/publish_release.py","scripts/verify_distribution.py","scripts/verify_lambda.sh","examples/library/guard_smoke.py",".github/workflows/ci.yml",".github/workflows/release.yml","docs/releases/README.md","tests/test_release.py"],"automatedChecks":[{"id":"release-safety","description":"Reject mismatched tags, altered assets and mutable publication.","command":"python -m pytest -q tests/test_release.py"},{"id":"installed-artifacts","description":"Install wheel and sdist outside the checkout and invoke public APIs.","command":"python scripts/verify_distribution.py --python 3.12"},{"id":"release-inventory","description":"Check archive contents, version metadata and SHA-256 inventory.","command":"python scripts/release.py verify"},{"id":"linux-offline-consumer","description":"Bundle and run with networking disabled in Linux Python 3.12.","command":"bash scripts/verify_lambda.sh"}],"manualChecks":["Follow the release guide and confirm Guard retains domain severity and incident ownership; inspect readiness and evidence limits before adopting."],"verificationGaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for interfaces/python/install-versioned-library.","criteria":[{"criterion":"Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","status":"partial","evidence":["src/anomalyzer/_version.py:1","scripts/release.py:1","scripts/release.py:20","scripts/release.py:39","scripts/release.py:48","scripts/release.py:54","scripts/publish_release.py:19","scripts/publish_release.py:27","scripts/publish_release.py:29","scripts/publish_release.py:39","scripts/verify_lambda.sh:14","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19","scripts/verify_lambda.sh:25","scripts/verify_lambda.sh:27",".github/workflows/ci.yml:14",".github/workflows/ci.yml:19",".github/workflows/ci.yml:27",".github/workflows/ci.yml:41",".github/workflows/ci.yml:46",".github/workflows/release.yml:28","docs/releases/README.md:12","docs/releases/README.md:13","docs/releases/README.md:14","docs/releases/README.md:17","docs/releases/README.md:18","tests/test_release.py:60","tests/test_release.py:71","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","status":"partial","evidence":["scripts/verify_distribution.py:1","scripts/verify_distribution.py:42","examples/library/guard_smoke.py:32","examples/library/guard_smoke.py:49","examples/library/guard_smoke.py:65","examples/library/guard_smoke.py:69","examples/library/guard_smoke.py:75",".github/workflows/ci.yml:57","docs/releases/README.md:13","docs/releases/README.md:208","docs/releases/README.md:211","docs/releases/README.md:213"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:34","scripts/release.py:81","scripts/release.py:82","scripts/release.py:85","scripts/release.py:92","scripts/publish_release.py:1","scripts/publish_release.py:25","scripts/publish_release.py:26","scripts/publish_release.py:27","scripts/publish_release.py:28","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:24","examples/library/guard_smoke.py:39",".github/workflows/ci.yml:42",".github/workflows/ci.yml:45",".github/workflows/ci.yml:46",".github/workflows/ci.yml:49",".github/workflows/ci.yml:51",".github/workflows/release.yml:31",".github/workflows/release.yml:33",".github/workflows/release.yml:36",".github/workflows/release.yml:37","docs/releases/README.md:3","docs/releases/README.md:12","docs/releases/README.md:14","docs/releases/README.md:15","docs/releases/README.md:19","tests/test_release.py:1","tests/test_release.py:21","tests/test_release.py:23","tests/test_release.py:52","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:81","scripts/release.py:82","scripts/publish_release.py:40","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:6","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19",".github/workflows/ci.yml:30",".github/workflows/ci.yml:31",".github/workflows/ci.yml:51",".github/workflows/ci.yml:52",".github/workflows/ci.yml:90","docs/releases/README.md:12","docs/releases/README.md:15","docs/releases/README.md:58","docs/releases/README.md:59","docs/releases/README.md:75"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","partial: Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","partial: Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","partial: A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","name":"Python integration","label":"Python integration / Embed analysis in my application / Install a released dependency"},"impact":0,"gaps":1,"r":41,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","label":"Ask an agent to explain un...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","acceptance":["The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit."],"guidance":["This contract covers the supplied instructions and demonstrated analysis workflow; it does not guarantee that every agent host follows the instructions.","Synthetic examples demonstrate mechanics, not real-world detection accuracy."],"path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/agents/openai.yaml","src/anomalyzer/cli.py","src/anomalyzer/mcp.py","tests/test_cli.py","tests/test_agentic.py","tests/test_mcp.py"],"automatedChecks":[{"id":"structured-cli","description":"Exercise the documented structured CLI workflow and preserve incomplete-history evidence.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence tests/test_cli.py::test_bare_json_values_cli"},{"id":"declared-relationships","description":"Check explicit ratio semantics and unavailable source evidence.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"mcp-drilldown","description":"Follow compact analysis to a finding and its source evidence.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case"}],"manualChecks":["Follow the conversion example through the CLI and MCP drilldown; compare the Skill instructions with actual evidence and check that its proposed text separates calibrated findings, early candidates, readiness, and caller-owned actions."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","criteria":[{"criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:13",".agents/skills/anomaly-detection/SKILL.md:41","tests/test_cli.py:13"],"notes":"Executable resolution and JSON commands are explicit. Manual conversion and spike CLI runs use the installed numerical engine."},{"criterion":"The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:24",".agents/skills/anomaly-detection/SKILL.md:32","tests/test_cli.py:51","tests/test_agentic.py:211"],"notes":"Instructions preserve scope, units, missingness, declared relationships, and exact alignment; existing runtime checks establish these prerequisites."},{"criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:101",".agents/skills/anomaly-detection/SKILL.md:134",".agents/skills/anomaly-detection/agents/openai.yaml:4"],"notes":"The supplied instructions lead with plain-language changes and then qualify maturity/readiness/overlap. This establishes instructional content, not measured agent compliance."},{"criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:54",".agents/skills/anomaly-detection/SKILL.md:174","src/anomalyzer/mcp.py:169","tests/test_mcp.py:54"],"notes":"The documented three-tool workflow matches executed conversion drilldown, with session-only run IDs and explicit read-only/caller-owned delivery limits."}],"ignore_findings":[],"done":true,"input_fingerprint":"751723214aa340b1227ec347c22f130749fcdbfd49327c21b4c27ef68d7f0341","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Get an evidence-grounded explanation"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","label":"See the evidence in a char...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","acceptance":["The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes."],"guidance":["Native chart rendering and interactive figure features depend on the agent host; the checked-in fallback produces a static SVG for a single dataset.","A chart explains structured analysis; it does not replace the detector result."],"path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/scripts/quick_chart.py",".agents/skills/anomaly-detection/references/visualization.md","tests/test_skill_chart.py"],"automatedChecks":[{"id":"observed-evidence","description":"Render real observed and expected data and only actual point triggers.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_preserves_real_point_evidence"},{"id":"maturity-and-resets","description":"Keep provisional evidence non-triggering and label reviewed regimes.","command":"python -m pytest -q tests/test_skill_chart.py::test_provisional_scores_are_not_charted_as_triggers tests/test_skill_chart.py::test_quick_chart_marks_reviewed_resets"},{"id":"unsupported-scope","description":"Reject a multi-dataset result rather than mixing units.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_rejects_multiple_datasets"},{"id":"output-protection","description":"Require an explicit overwrite for existing charts.","command":"python -m pytest -q tests/test_skill_chart.py::test_chart_cli_requires_explicit_overwrite"}],"manualChecks":["Inspect a spike fallback chart and the detailed figure instructions; verify that axes, initialization context, trigger markers, and interpretation limits are clear, and that the Skill requests a chart only when the user asks."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","criteria":[{"criterion":"The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:189",".agents/skills/anomaly-detection/SKILL.md:226"],"notes":"The Skill explicitly requires a visual request after analysis and separates quick versus detailed rendering instructions."},{"criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:81",".agents/skills/anomaly-detection/scripts/quick_chart.py:59","tests/test_skill_chart.py:28","tests/test_skill_chart.py:43","tests/test_skill_chart.py:54"],"notes":"Real-result tests establish observed/expected lines, trigger-only markers, provisional non-triggering, units, initialization shading, and reset boundaries; the spike chart was inspected visually."},{"criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:85",".agents/skills/anomaly-detection/scripts/quick_chart.py:251","tests/test_skill_chart.py:63","tests/test_skill_chart.py:68"],"notes":"Multiple resolved datasets raise a specific error. The CLI test preserves an existing chart and replaces it only with explicit overwrite."},{"criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/references/visualization.md:31",".agents/skills/anomaly-detection/references/visualization.md:37",".agents/skills/anomaly-detection/references/visualization.md:45",".agents/skills/anomaly-detection/references/visualization.md:86"],"notes":"Detailed instructions require index alignment, maturity distinctions, first detection timing, visible unavailable status, and qualified causal claims; native interactive rendering remains host-dependent."}],"ignore_findings":[],"done":true,"input_fingerprint":"2780b6c3e363f89a30a0709772c0a46a9a96f1567fabb27e63934a3ca80dc982","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Inspect a requested chart"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","label":"Find which segments contri...","area":"investigation","scopes":["investigation","investigation/cohorts"],"scope":"investigation/cohorts","scopeLabel":"investigation/cohorts","status":"planned","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","acceptance":["Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","Segment ranking retains absolute signed contributions and baseline support; an arithmetic contributor is not automatically labeled anomalous or causal."],"guidance":["Ratios and overlapping customer groups are not additive totals; separate segment expectations must not be assumed to sum to a fitted total expectation.","Proposed automated checks cover exact reconciliation, offsetting segment changes, missing segments, and zero total departure.","Manual review should inspect a regional decline with an offsetting increase and confirm the explanation preserves signed contributions and evidence limits."],"path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Locate contributors to a total change"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","label":"Keep reviewed outcomes bes...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","acceptance":["A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","Review export distinguishes analyst judgment from detector output; annotation alone does not change calibration, eligibility, or regime boundaries."],"guidance":["Caller-supplied reviewer identity is attribution, not built-in authentication; storage, access control, response execution, and incident ownership stay with callers.","Proposed automated checks cover evidence-version linking, unresolved reviews, revision retention, and unchanged numerical evidence after annotation.","Manual review should inspect a disputed case with two review revisions and confirm judgments, uncertainties, and original findings remain distinguishable."],"path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Record the reviewed outcome"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","label":"Find earlier reviewed case...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","acceptance":["Search is bounded to caller-selected records and explicit compatibility filters; it does not scan unrelated local files or require a model API.","Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match."],"guidance":["Similar historical evidence does not establish the same cause or authorize repeating a past response; caller-selected records may be incomplete or biased.","Proposed automated checks cover compatibility filtering, bounded corpus access, missing comparison context, and no-match behavior.","Manual review should compare superficially similar cases with different outcomes and confirm precedent is presented as investigation context."],"path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Inspect comparable reviewed history"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","label":"Change action eligibility ...","area":"investigation","scopes":["investigation","investigation/policy"],"scope":"investigation/policy","scopeLabel":"investigation/policy","status":"implemented","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","acceptance":["Default policy leaves action and notification eligibility disabled.","Explicit-rule and calibrated-departure eligibility are controlled independently.","Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged."],"guidance":["Eligibility does not send notifications, execute actions, check readiness, or confer authority to act.","Monitoring schedules, delivery, reviewed incident state, and deduplication belong to the caller."],"path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","dependencies":["investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/agent.py","src/anomalyzer/cases.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"independent-notifications","description":"Replay notification and action choices without changing numerical evidence or the original result.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_separates_action_from_notification"},{"id":"calibrated-choice","description":"Enable calibrated departures only with the corresponding policy choice.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_enables_calibrated_departures_only_when_requested"},{"id":"nontriggering-candidates","description":"Keep descriptive candidates ineligible even under a permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"}],"manualChecks":["Inspect the unit-cost example under default and allow_explicit_rules policies and compare eligibility with the unchanged numerical violation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","criteria":[{"criterion":"Default policy leaves action and notification eligibility disabled.","status":"covered","evidence":["src/anomalyzer/contracts.py:329","tests/test_capability_behaviors.py:74"],"notes":"Default flags disable both eligibility types, including after replay resets policy."},{"criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"covered","evidence":["src/anomalyzer/agent.py:47","tests/test_capability_behaviors.py:123"],"notes":"Explicit-only permission does not enable a calibrated departure; the matching calibrated choice does."},{"criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"covered","evidence":["src/anomalyzer/agent.py:53","tests/test_capability_behaviors.py:101","tests/test_capability_behaviors.py:134"],"notes":"Notification-only replay enables a met rule while keeping action disabled; permissive settings still cannot enable candidates."},{"criterion":"Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged.","status":"covered","evidence":["src/anomalyzer/agent.py:42","src/anomalyzer/cases.py:57","src/anomalyzer/cases.py:105","tests/test_capability_behaviors.py:101","tests/test_agentic.py:49"],"notes":"Assertions compare unchanged original serialization, methods, evidence references, stable IDs, and eligibility-dependent revision changes."}],"ignore_findings":[],"done":true,"input_fingerprint":"ee67bcc1ec1380b8d84a0556d838d648ae6cef3f4f9b9e34e1b71e90eb00aeae","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Review action eligibility"},"impact":0,"gaps":0,"r":38,"x":1110.6666666666667,"y":760},{"id":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","label":"Review a possible new base...","area":"investigation","scopes":["investigation","investigation/regimes"],"scope":"investigation/regimes","scopeLabel":"investigation/regimes","status":"planned","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","acceptance":["A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","Producing a proposal leaves the active regime and original findings unchanged.","A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect."],"guidance":["A proposed boundary is not a confirmed cause or a retrospectively available detection; adaptive reference updates remain a separate behavior.","Proposed automated checks cover unchanged active results, short new regimes, rejected proposals, and accepted explicit-reset equivalence.","Manual review should inspect a persistent outage-like drop and a known launch and confirm neither becomes ordinary behavior merely because it persists."],"path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Review a proposed regime transition"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","label":"Distinguish changing perfo...","area":"relationships","scopes":["relationships","relationships/cohorts"],"scope":"relationships/cohorts","scopeLabel":"relationships/cohorts","status":"planned","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","acceptance":["Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","The report preserves aggregate and cohort rates, population weights, and the declared decomposition formula and reference-period convention.","Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes."],"guidance":["Decomposition is descriptive accounting; it does not estimate an intervention effect or establish that the comparison populations are causally equivalent.","Proposed automated checks cover stable rates with changing mix, stable mix with changing rates, offsetting cohort rates, and new or empty cohorts.","Manual review should inspect a Simpson's-paradox example and confirm the aggregate movement and cohort evidence can both be understood correctly."],"path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Separate performance changes from population mix"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878},{"id":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","label":"Find changes in conversion...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","acceptance":["Ratios retain the computed value, formula, and original source indexes for each sample.","Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","Derived series can be analyzed with the frozen or adaptive single-resolution recipe."],"guidance":["Relationships are caller-supplied declarations, not discovered or causal relationships.","A difference requires identical declared units; a normalized residual requires a positive supplied scale.","A joint condition is a declared 0/1 comparison, not statistical proof."],"path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","dependencies":["relationships/quality/require-compatible-sources","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_examples.py"],"automatedChecks":[{"id":"ratio-lineage","description":"Explain a changing rate with the exact source indexes and formula.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"other-comparisons","description":"Check declared differences, normalized residuals, and combined conditions.","command":"python -m pytest -q tests/test_agentic.py::test_remaining_relationship_primitives"},{"id":"unavailable-values","description":"Keep zero-denominator and missing-source observations unavailable.","command":"python -m pytest -q tests/test_agentic.py::test_missing_and_zero_denominator_are_unavailable_not_zero"},{"id":"adaptive-relationship","description":"Use the adaptive recipe for a declared relationship.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_recipe_runs_for_declared_relationships"}],"manualChecks":["Inspect examples/agentic/conversion.json and distinguish its relationship departure from the individual source metrics."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","criteria":[{"criterion":"Ratios retain the computed value, formula, and original source indexes for each sample.","status":"covered","evidence":["src/anomalyzer/relationships.py:84","src/anomalyzer/relationships.py:152","tests/test_agentic.py:211"],"notes":"Assertions check exact ratio values, formula, and both original source indexes."},{"criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"covered","evidence":["src/anomalyzer/relationships.py:97","src/anomalyzer/relationships.py:134","tests/test_agentic.py:285"],"notes":"Parameterized assertions independently specify difference, normalized residual, all-condition, and any-condition outputs."},{"criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"covered","evidence":["src/anomalyzer/relationships.py:89","src/anomalyzer/relationships.py:142","tests/test_agentic.py:226"],"notes":"Lineage retains null values with missing_source_value/zero_denominator reasons. Explicit zero_denominator=error intentionally fails instead, as documented."},{"criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:87","src/anomalyzer/orchestrator.py:135","tests/test_agentic.py:171","tests/test_examples.py:17"],"notes":"Both supported recipes dispatch derived datasets through the shared single-resolution engine."}],"ignore_findings":[],"done":true,"input_fingerprint":"7fa8f7ffdc0d852ad961b3e1753761d1bd3b00b87b8c001a957ec19e26d068b9","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Compare ratios and metric gaps"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","label":"Check whether a metric fol...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","acceptance":["A supplied lag and coefficient use the predictor at the declared earlier sample position.","Derived values retain response and predictor source indexes and source cutoffs in lineage.","When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","Samples lacking the earlier predictor or awaiting coefficient training remain unavailable."],"guidance":["Lag is measured in samples and is never searched automatically.","Exactly one of coefficient or training_size is required."],"path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","dependencies":["relationships/quality/require-compatible-sources"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_examples.py","examples/agentic/lagged-response.json"],"automatedChecks":[{"id":"supplied-response","description":"Use the declared lag and coefficient with correct predictor indexes.","command":"python -m pytest -q tests/test_agentic.py::test_lagged_response_uses_supplied_coefficient_and_prior_sample"},{"id":"frozen-coefficient","description":"Keep fitted coefficient and earlier lineage invariant when future samples arrive.","command":"python -m pytest -q tests/test_agentic.py::test_estimated_lag_coefficient_is_frozen_and_prefix_invariant"},{"id":"delayed-example","description":"Run the checked-in lagged-response demonstration against expected results.","command":"python -m pytest -q tests/test_examples.py::test_agentic_demo_checks_all_cases"},{"id":"pending-training","description":"Retain unavailable lineage until the full declared training prefix arrives, then freeze its coefficient.","command":"python -m pytest -q tests/test_agentic.py::test_lag_fit_waits_for_complete_training_prefix"}],"manualChecks":["Inspect the lagged-response example's leading unavailable values, coefficient parameters, and first usable derived sample."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","criteria":[{"criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"covered","evidence":["src/anomalyzer/relationships.py:115","tests/test_agentic.py:254"],"notes":"Supplied lag/coefficient uses the declared earlier predictor position and preserves the expected response residual."},{"criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"covered","evidence":["src/anomalyzer/relationships.py:152","tests/test_agentic.py:254","examples/agentic/lagged-response.json"],"notes":"Available and pending lineage retains existing source indexes and timestamp cutoffs; missing predictors do not become zero."},{"criterion":"When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","status":"covered","evidence":["src/anomalyzer/relationships.py:47","tests/test_agentic.py:294","tests/test_agentic.py:330"],"notes":"The coefficient is fitted once after the complete declared prefix; the independent 20/7 result and prefix-invariance assertions pass."},{"criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"covered","evidence":["src/anomalyzer/relationships.py:122","tests/test_agentic.py:330"],"notes":"Short prefixes retain null derived values and explicit unavailable reasons, while an unidentifiable completed fit has a visible error."}],"ignore_findings":[],"done":true,"input_fingerprint":"fc4516ef6ab903dfa4918268d97049aff04fbf9556620ccb8259ff830ba85785","mode":"quick"},"reviewFindings":[],"reviewStale":false,"health":"implemented","healthLabel":"Implemented","coverage":"full","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Check delayed responses"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","label":"Inspect related metrics be...","area":"relationships","scopes":["relationships","relationships/resolution"],"scope":"relationships/resolution","scopeLabel":"relationships/resolution","status":"planned","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed calendar or fixed-duration aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","acceptance":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"guidance":["Initial support should use explicit relationships and compatible boundaries; cross-cadence joining or automatic resampling needs a separate contract.","Proposed automated checks cover ratio-of-totals arithmetic, missing sources, mismatched period boundaries, and open-period evidence separation.","Manual review should inspect a conversion decline within an unfinished day and verify the explanation separates native evidence from partial totals."],"path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/resolution/compare-finalized-subperiods has no automated or manual verification checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/resolution/compare-finalized-subperiods has no automated checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-manual-review","message":"relationships/resolution/compare-finalized-subperiods has no manual review guidance","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Compare related metrics while the period is open"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878}],"links":[{"source":"data/inputs/accept-analysis-requests","target":"analysis/evidence/calibrate-departures","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/cli/analyze-files-and-streams","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"analysis/expectations/model-seasonal-level-and-growth","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"interfaces/python/analyze-through-public-api","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"investigation/assessments/describe-early-changes","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"relationships/quality/require-compatible-sources","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/identify-residual-patterns","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"data/quality/exclude-incomplete-periods","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/mcp/connect-read-only-local-server","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/rules/check-declared-numerical-rules","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/report-point-anomaly-episodes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/evidence/assess-reference-readiness","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/adapt-to-changing-growth","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/protect-reference-from-extremes","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"data/quality/exclude-incomplete-periods","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"execution/resilience/retain-evidence-under-failure-and-limits","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/mcp/connect-read-only-local-server","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/visualize-analysis-on-request","kind":"committed"},{"source":"investigation/rules/check-declared-numerical-rules","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/evaluate-delayed-response","kind":"committed"}],"scopes":["analysis","analysis/detection","analysis/evidence","analysis/expectations","analysis/rates","analysis/resolution","data","data/calendar","data/cohorts","data/inputs","data/quality","data/revisions","evaluation","evaluation/comparisons","evaluation/outcomes","evaluation/replay","execution","execution/incremental","execution/resilience","execution/state","interfaces","interfaces/cli","interfaces/export","interfaces/mcp","interfaces/python","interfaces/skill","investigation","investigation/assessments","investigation/cases","investigation/cohorts","investigation/history","investigation/policy","investigation/regimes","investigation/rules","relationships","relationships/cohorts","relationships/comparisons","relationships/quality","relationships/resolution"],"storyMaps":["Anomaly detection Skill","Basic CLI anomaly detection","Evidence and review policy","Future - Business-aware expectations","Future - Calendar and data readiness","Future - Cohort investigation","Future - Continuous analysis","Future - Historical evaluation","Future - Investigation memory","Incomplete-period monitoring","Python integration","Read-only MCP server","Related-metric analysis","Trend and seasonal analysis"],"hasUnassignedStoryMap":false};
+const graph = {"nodes":[{"id":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","label":"Understand how much histor...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","acceptance":["Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","Provisional residual scores can be reported but do not trigger point-anomaly episodes.","Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger."],"guidance":["Evidence strength is not a probability of an incident.","Calibration and expectations are separate concerns; each declared regime calibrates again."],"path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/detection/identify-residual-patterns","analysis/detection/report-point-anomaly-episodes","analysis/evidence/assess-reference-readiness","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","relationships/comparisons/compare-declared-metrics"],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"short-history","description":"Keep reference-only evidence for a short series.","command":"python -m pytest -q tests/test_analysis.py::test_short_series_returns_reference_evidence"},{"id":"provisional-evidence","description":"Report provisional scores without point triggers.","command":"python -m pytest -q tests/test_analysis.py::test_early_evidence_maturity_and_no_provisional_trigger"},{"id":"independent-calibration","description":"Check a known residual center, sample scale, and standardized score.","command":"python -m pytest -q tests/test_analysis.py::test_calibration_reference"},{"id":"constant-reference","description":"Handle zero residual variance without spurious episodes.","command":"python -m pytest -q tests/test_analysis.py::test_constant_and_schema"}],"manualChecks":["Inspect a sample's signal_maturity, calibration_samples, standardized_residual, and triggers together before describing a statistical departure."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","criteria":[{"criterion":"Short series return reference evidence and an insufficient_history status instead of inventing calibrated findings.","status":"covered","evidence":["src/anomalyzer/recipes.py:205","src/anomalyzer/recipes.py:445","tests/test_analysis.py:141"],"notes":"The short-series test checks three reference-only samples, absent standardized scores, and no triggers."},{"criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"covered","evidence":["src/anomalyzer/recipes.py:210","tests/test_analysis.py:123"],"notes":"A provisional score of 2 exceeds the configured point threshold of 1 but produces no trigger or episode."},{"criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"covered","evidence":["src/anomalyzer/recipes.py:231","src/anomalyzer/recipes.py:307","tests/test_analysis.py:111"],"notes":"Calibration uses retained residual mean and sample standard deviation once per regime; the independent numerical example produces the expected standardized score."},{"criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"covered","evidence":["src/anomalyzer/contracts.py:104","src/anomalyzer/recipes.py:350","src/anomalyzer/recipes.py:388","tests/test_analysis.py:111"],"notes":"Sample fields retain the comparison and support; the point threshold is inspectable in the enclosing method parameters."}],"ignore_findings":[],"done":true,"input_fingerprint":"ccdc1b2e92f3e647d5a9e6127cc3c4b10bf0da8f1f78af3efb805daa9b1e4305","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Establish evidence support"},"impact":6,"gaps":0,"r":68,"x":170,"y":170},{"id":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","label":"Bring my metrics into an a...","area":"data","scopes":["data","data/inputs"],"scope":"data/inputs","scopeLabel":"data/inputs","status":"implemented","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","acceptance":["Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","Published schema 1.0 and 1.1 request and result schemas match the public Python contracts."],"guidance":["Null observations remain unknown unless a schema-1.1 additive-volume policy explicitly authorizes confirmed empty collection as zero. Limits apply to the request as a whole.","Describe current supported behavior; historical plans do not add accepted inputs."],"path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["analysis/evidence/calibrate-departures","analysis/expectations/model-seasonal-level-and-growth","data/quality/analyze-sparse-observations","execution/resilience/retain-evidence-under-failure-and-limits","interfaces/python/analyze-through-public-api","investigation/assessments/describe-early-changes"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_agentic.py","schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json"],"automatedChecks":[{"id":"input-forms","description":"Analyze a bare list and validate a canonical request and result.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings tests/test_analysis.py::test_constant_and_schema"},{"id":"explicit-scope","description":"Reject schema 1.1 datasets without entity scope.","command":"python -m pytest -q tests/test_agentic.py::test_schema_11_requires_explicit_entity_scope"},{"id":"invalid-inputs","description":"Reject unsupported settings and nonnumeric or nonfinite observations.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_config tests/test_analysis.py::test_invalid_values"},{"id":"published-schemas","description":"Compare all four checked-in schemas with their public contracts.","command":"python -m pytest -q tests/test_analysis.py::test_checked_in_schemas tests/test_agentic.py::test_v11_checked_in_schemas_exist"}],"manualChecks":["Review the input examples in README.md and confirm they describe the two supported schema versions and the four-dataset/four-relationship limit."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","criteria":[{"criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"covered","evidence":["src/anomalyzer/recipes.py:21","src/anomalyzer/contracts.py:96","tests/test_analysis.py:234"],"notes":"Bare-list settings are validated; schema 1.0 limits datasets to exactly one."},{"criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"covered","evidence":["src/anomalyzer/contracts.py:240","src/anomalyzer/contracts.py:335","tests/test_agentic.py:192"],"notes":"Entity scope is required, relationship kinds are discriminated, and references must name distinct known datasets."},{"criterion":"Unknown settings, unsupported recipes, nonfinite values, booleans, and numeric strings are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:7","src/anomalyzer/contracts.py:71","tests/test_analysis.py:526","tests/test_analysis.py:531"],"notes":"Assertions reject each listed category; missing null values remain a separate accepted representation."},{"criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"covered","evidence":["schemas/request.schema.json","schemas/result.schema.json","schemas/request-1.1.schema.json","schemas/result-1.1.schema.json","tests/test_analysis.py:547","tests/test_agentic.py:317"],"notes":"Tests and independent inspection compare each complete published schema with its contract-generated schema. Python cross-field validators supplement JSON Schema constraints."}],"ignore_findings":[],"done":true,"input_fingerprint":"3546cd8f436812ff11c9352a855926677b55cc6287e40f6a5041bd613173e5b4","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Supply a numeric series"},"impact":6,"gaps":0,"r":68,"x":353.33333333333337,"y":170},{"id":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","label":"Analyze local files from t...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","acceptance":["The methods command lists supported analysis recipes and residual detectors.","Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status."],"guidance":["CSV callers must select one entity and unit before analysis.","Exit 0 includes completed analysis with anomalies, insufficient history, or inapplicability; exit 2 means invalid input, and exit 1 means execution failure or partial execution."],"path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/python/install-versioned-library","interfaces/skill/guide-anomaly-investigation","interfaces/skill/visualize-analysis-on-request"],"implementationReferences":["src/anomalyzer/cli.py","src/anomalyzer/io.py","src/anomalyzer/__main__.py","tests/test_cli.py"],"automatedChecks":[{"id":"file-stdin","description":"Analyze equivalent file and stdin requests.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence"},{"id":"csv-precedence","description":"Validate selected CSV fields and apply flag-over-file-over-embedded precedence.","command":"python -m pytest -q tests/test_cli.py::test_csv_and_precedence"},{"id":"value-list","description":"Analyze a bare JSON array and expose inferred settings.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"},{"id":"output-and-exits","description":"Protect existing output and distinguish invalid options from failed execution.","command":"python -m pytest -q tests/test_cli.py::test_output_protection tests/test_cli.py::test_methods_and_invalid_options tests/test_cli.py::test_size_limit_and_execution_exit"}],"manualChecks":["Run an example in readable and JSON modes and confirm stdout contains results while diagnostics go to stderr."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","criteria":[{"criterion":"The methods command lists supported analysis recipes and residual detectors.","status":"covered","evidence":["src/anomalyzer/cli.py:426","src/anomalyzer/registry.py:15","tests/test_cli.py:27"],"notes":"Discovery tests check all three recipes and ten residual detectors, with structured stdout and no successful-run diagnostics on stderr."},{"criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:239","tests/test_cli.py:13","tests/test_mcp.py:163"],"notes":"The passing file/stdin regression compares request fingerprints and short-history semantics; cross-interface tests compare complete stable numerical output."},{"criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"covered","evidence":["src/anomalyzer/io.py:33","tests/test_cli.py:51"],"notes":"The passing CSV regression covers timestamped missing cells, positional data, and varying unmapped entity rejection."},{"criterion":"Settings resolve in embedded-request, settings-file, then CLI-flag precedence; bare JSON values can use CLI configuration.","status":"covered","evidence":["src/anomalyzer/cli.py:253","tests/test_cli.py:51","tests/test_cli.py:71"],"notes":"The passing file/flag/embedded precedence regression and bare-values regression establish the merge order and validated configuration."},{"criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"covered","evidence":["src/anomalyzer/io.py:69","src/anomalyzer/cli.py:424","tests/test_cli.py:41","tests/test_cli.py:108"],"notes":"Output protection and exit regressions passed, including explicit overwrite; successful manual examples have empty stderr and incomplete-usage exits 1."}],"ignore_findings":[],"done":true,"input_fingerprint":"86e701ccc0a1aefd407f11b4547784fef497fa1b49ddf6bb69db0d08f757dfdc","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Run from files or stdin"},"impact":4,"gaps":0,"r":58,"x":903.3333333333334,"y":170},{"id":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","label":"Account for ordinary seaso...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","acceptance":["A supplied season length selects earlier matching seasonal observations for the reference.","Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected."],"guidance":["Expectations support scoring; they are not a standalone forecasting API."],"path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["analysis/expectations/protect-reference-from-extremes","analysis/expectations/reset-declared-regimes","analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py"],"automatedChecks":[{"id":"seasonal-reference","description":"Check independent lagged-reference values.","command":"python -m pytest -q tests/test_analysis.py::test_forecast_independent_reference"},{"id":"growth-and-departures","description":"Keep clean linear and compound growth ordinary while identifying injected departures.","command":"python -m pytest -q tests/test_trend.py::test_trend_and_anomalies"},{"id":"frozen-fit","description":"Keep the training fit stable when later observations change.","command":"python -m pytest -q tests/test_trend.py::test_frozen_fit_and_legacy_mode"},{"id":"explicit-trends","description":"Accept supported trend choices and handle short, constant, and signed series.","command":"python -m pytest -q tests/test_trend.py::test_explicit_modes_and_validation tests/test_trend.py::test_short_history_constant_and_signed_linear"}],"manualChecks":["Compare examples/growth_up.json and examples/compound.json with their selected trend and evidence expectations."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","criteria":[{"criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"covered","evidence":["src/anomalyzer/baseline.py:8","src/anomalyzer/baseline.py:31","tests/test_analysis.py:21"],"notes":"Independent lag values and robust matching-phase assertions establish the seasonal reference calculation."},{"criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"covered","evidence":["src/anomalyzer/trend.py:120","src/anomalyzer/trend.py:144","tests/test_trend.py:11"],"notes":"Positive and negative linear/exponential cases select the expected model, match post-training expectations, and produce no ordinary-growth episodes."},{"criterion":"The fitted trend remains frozen after training within a regime; later changed values do not rewrite that fit.","status":"covered","evidence":["src/anomalyzer/recipes.py:162","tests/test_trend.py:36"],"notes":"Trend fitting occurs at the training boundary; the test asserts identical trend diagnostics and earlier evidence after later observations change."},{"criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:20","tests/test_trend.py:49","tests/test_trend.py:57"],"notes":"Tests exercise disabled, linear, exponential, and invalid quadratic modes, including disclosed short-history fallback."}],"ignore_findings":[],"done":true,"input_fingerprint":"8e9a0361b02b7d3f651366e5fa0675cccb3949e44de448d2ae0d43e6710bd408","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Account for growth and seasonality"},"impact":3,"gaps":0,"r":53,"x":170,"y":170},{"id":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","label":"Embed local analysis in my...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","acceptance":["analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","analyze_series preserves supplied settings and event-time context for a single dataset.","analyze_relationships preserves the declared multi-dataset relationship semantics.","Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime."],"guidance":["All interfaces use local analysis; there is no hosted service or model API requirement."],"path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence","interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/__init__.py","src/anomalyzer/agent.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"request-parity","description":"Match canonical request results across Python, CLI, and full-result MCP compatibility calls.","command":"python -m pytest -q tests/test_mcp.py::test_canonical_request_matches_python_cli_and_mcp tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"series-context","description":"Preserve a single-series event-time cutoff and context across interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp tests/test_mcp.py::test_series_wrapper_preserves_omitted_and_explicit_season_settings"},{"id":"relationship-parity","description":"Match declared relationship evidence across the public interfaces.","command":"python -m pytest -q tests/test_mcp.py::test_relationship_semantics_match_python_cli_and_mcp"}],"manualChecks":["Review the Python API examples in README.md and confirm callers can inspect the returned validated models and JSON."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","criteria":[{"criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"covered","evidence":["src/anomalyzer/__init__.py:3","src/anomalyzer/recipes.py:21","tests/test_analysis.py:234"],"notes":"Bare values and separate settings validate before dispatch; canonical schema 1.0 and 1.1 requests use their respective contracts."},{"criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"covered","evidence":["src/anomalyzer/agent.py:11","tests/test_mcp.py:191","tests/test_mcp.py:291"],"notes":"Field-set handling preserves omitted versus explicit settings and context in the single-series wrapper, verified across dictionary and Settings inputs."},{"criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"covered","evidence":["src/anomalyzer/agent.py:27","tests/test_mcp.py:217"],"notes":"Declared relationships validate and use the shared engine without changing units, lineage, or primitive semantics."},{"criterion":"Equivalent requests retain numerical results across Python, CLI, and full-result MCP compatibility calls apart from run IDs and runtime.","status":"covered","evidence":["tests/test_mcp.py:163","tests/test_mcp.py:191","tests/test_mcp.py:217","src/anomalyzer/__main__.py:1"],"notes":"Passing real CLI subprocess, Python, and full-result MCP parity tests compare stable numerical results apart from identity and runtime."}],"ignore_findings":[],"done":true,"input_fingerprint":"214bd10bb3aa5fad0f74938fe2ef09853a593cdb5b93cba49ec851a628d8ab65","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","name":"Python integration","label":"Python integration / Embed analysis in my application / Call the public API"},"impact":3,"gaps":0,"r":53,"x":903.3333333333334,"y":170},{"id":"investigation/cases/group-related-evidence","title":"Investigate related findings together","label":"Investigate related findin...","area":"investigation","scopes":["investigation","investigation/cases"],"scope":"investigation/cases","scopeLabel":"investigation/cases","status":"implemented","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","acceptance":["Evidence for different declared entities remains in separate cases even at the same event time.","Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","A derived relationship adds context without counting as an additional independent source for severity.","Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection."],"guidance":["Warning and critical are deterministic numerical grouping rules, not validated incident impact.","Diagnostic trend, oscillation, mixture, and variation findings remain distinct from adaptive location-shift cases."],"path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","dependencies":["investigation/assessments/describe-early-changes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["interfaces/cli/explain-findings-in-readable-output","interfaces/mcp/investigate-findings-and-evidence","investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/cases.py","src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py"],"automatedChecks":[{"id":"entity-scope","description":"Separate cases for different entities at the same event time.","command":"python -m pytest -q tests/test_agentic.py::test_cases_are_scoped_by_entity_at_the_same_event_time"},{"id":"source-severity","description":"Escalate numerical severity when two source datasets meet criteria.","command":"python -m pytest -q tests/test_agentic.py::test_two_source_datasets_escalate_case_severity"},{"id":"correlated-context","description":"Keep relationship evidence from double-counting source severity.","command":"python -m pytest -q tests/test_agentic.py::test_relationship_evidence_explains_but_does_not_double_count_severity"},{"id":"first-detection","description":"Group an adaptive location shift before point triggers without using a later pattern peak.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_location_shift_creates_one_early_case_without_point_trigger tests/test_agentic.py::test_location_case_severity_does_not_use_future_pattern_peak"}],"manualChecks":["Inspect a case's contributing targets, explanation, evidence references, and investigation questions; distinguish it from a point-anomaly episode."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","criteria":[{"criterion":"Evidence for different declared entities remains in separate cases even at the same event time.","status":"covered","evidence":["src/anomalyzer/cases.py:42","tests/test_agentic.py:69"],"notes":"Entity participates in grouping and stable identity; the test requires separate cases and distinct IDs at the same time."},{"criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"covered","evidence":["src/anomalyzer/cases.py:72","tests/test_agentic.py:85"],"notes":"Two qualifying source datasets produce critical numerical severity and named source explanations."},{"criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"covered","evidence":["src/anomalyzer/cases.py:72","src/anomalyzer/cases.py:116","tests/test_agentic.py:100"],"notes":"The independent-source count includes dataset criteria only. Derived evidence remains explanatory and can still meet the separately declared magnitude rule."},{"criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"covered","evidence":["src/anomalyzer/early.py:193","src/anomalyzer/early.py:237","src/anomalyzer/orchestrator.py:115","tests/test_agentic.py:122","tests/test_agentic.py:146"],"notes":"Tests require one moderate-shift case without point triggers and restrict assessment references/strength to samples at or before first detection."}],"ignore_findings":[],"done":true,"input_fingerprint":"6830af8a2f064123c12adb062f4793e446b97b3b805fd73a934ea84cd0d5d9cf","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Investigate related evidence"},"impact":3,"gaps":0,"r":53,"x":1086.6666666666667,"y":170},{"id":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","label":"Spot sustained shifts and ...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","acceptance":["Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","Trend, oscillation, and mixture findings retain their specific pattern meanings.","Moving-range findings identify variation changes separately from location changes."],"guidance":["Pattern findings may overlap and are not independent confirmations.","A residual location shift does not establish a change in the full distribution."],"path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["data/quality/analyze-sparse-observations","investigation/cases/group-related-evidence"],"implementationReferences":["src/anomalyzer/evidence.py","src/anomalyzer/recipes.py","tests/test_analysis.py"],"automatedChecks":[{"id":"location-shift","description":"Find moderate increases and decreases without point anomalies.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_patterns_describe_location_shift_without_point_anomaly"},{"id":"accumulating-change","description":"Detect a known CUSUM accumulation at the expected position.","command":"python -m pytest -q tests/test_analysis.py::test_cusum_detects_accumulated_moderate_location_shift"},{"id":"diagnostic-patterns","description":"Distinguish residual trend, oscillation, and mixture rules.","command":"python -m pytest -q tests/test_analysis.py::test_nelson_diagnostics_have_specific_semantics"},{"id":"variation","description":"Describe a moving-range increase as variation rather than location.","command":"python -m pytest -q tests/test_analysis.py::test_moving_range_detects_short_term_variation_increase"}],"manualChecks":["Compare examples/location_shift_up.json with its point episodes and pattern findings, allowing for overlapping residual rules."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","criteria":[{"criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"covered","evidence":["src/anomalyzer/evidence.py:118","src/anomalyzer/evidence.py:189","tests/test_analysis.py:293"],"notes":"Tests assert no point triggers or episodes and exact upward/downward detection positions for Nelson rules and CUSUM."},{"criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"covered","evidence":["src/anomalyzer/evidence.py:211","src/anomalyzer/evidence.py:230","tests/test_analysis.py:350"],"notes":"The known accumulation detects at index 16 and reports peak CUSUM approximately 6, threshold 5, and increase direction."},{"criterion":"Trend, oscillation, and mixture findings retain their specific pattern meanings.","status":"covered","evidence":["src/anomalyzer/evidence.py:70","src/anomalyzer/evidence.py:122","tests/test_analysis.py:340"],"notes":"Separate predicates and descriptions preserve the three meanings and explicitly limit location/distribution claims."},{"criterion":"Moving-range findings identify variation changes separately from location changes.","status":"covered","evidence":["src/anomalyzer/evidence.py:234","tests/test_analysis.py:360"],"notes":"Adjacent standardized-residual differences produce variation_shift findings; the test checks magnitude, detection index, and interpretation."}],"ignore_findings":[],"done":true,"input_fingerprint":"63716eb67db6b73964b79897d8c81f2891df2586ff213c95ea81c5699b8abbea","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Find sustained residual changes"},"impact":2,"gaps":0,"r":48,"x":170,"y":170},{"id":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","label":"Analyze observations in th...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","acceptance":["Calendar datasets provide timestamps and frequency; positional datasets provide neither. Explicit schema-1.1 event-index sources retain timestamps without a calendar frequency.","Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration."],"guidance":["No interpolation or implicit filling, resampling, or unit conversion occurs. Schema-1.1 source policies may explicitly authorize confirmed empty additive volume as zero.","Duplicate null values propagate through aggregation."],"path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":["data/quality/exclude-incomplete-periods","relationships/quality/require-compatible-sources"],"implementationReferences":["src/anomalyzer/sparse.py","tests/test_sparse.py","src/anomalyzer/validation.py","src/anomalyzer/contracts.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"coordinate-contract","description":"Reject partial timestamp metadata and positional as_of cutoffs.","command":"python -m pytest -q tests/test_analysis.py::test_timestamp_metadata_is_all_or_nothing"},{"id":"local-time","description":"Normalize valid local times and reject ambiguous or nonexistent times.","command":"python -m pytest -q tests/test_analysis.py::test_local_time"},{"id":"duplicates-cutoff","description":"Require an explicit duplicate policy and retain counts after the cutoff.","command":"python -m pytest -q tests/test_analysis.py::test_duplicate_sort_and_cutoff"},{"id":"cadence-context","description":"Check fixed weekly cadence and preserve cutoff semantics across interfaces.","command":"python -m pytest -q tests/test_analysis.py::test_weekly_cadence tests/test_mcp.py::test_single_series_context_matches_python_cli_and_mcp"}],"manualChecks":["Inspect a timestamped result's data_quality transformations, UTC window, and resolved context to confirm preparation is explainable."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","criteria":[{"criterion":"Timestamped datasets provide both timestamps and frequency; positional datasets provide neither.","status":"covered","evidence":["src/anomalyzer/contracts.py:77","tests/test_analysis.py:418"],"notes":"Partial coordinate metadata and positional as_of are rejected."},{"criterion":"Local timestamps require a timezone, and ambiguous or nonexistent daylight-saving times require an explicit offset.","status":"covered","evidence":["src/anomalyzer/validation.py:8","tests/test_analysis.py:500"],"notes":"UTC round trips detect ambiguous/nonexistent local times; valid local time normalizes to the expected UTC hour."},{"criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"covered","evidence":["src/anomalyzer/validation.py:87","src/anomalyzer/validation.py:103","tests/test_analysis.py:478"],"notes":"Explicit mean/sum handling retains duplicate counts and transformation descriptions. Nulls propagate rather than becoming zero."},{"criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"covered","evidence":["src/anomalyzer/validation.py:105","src/anomalyzer/validation.py:111","tests/test_analysis.py:491","tests/test_mcp.py:191"],"notes":"Filtering is inclusive through as_of; cadence compares UTC elapsed timedeltas, including fixed seven-day weeks."}],"ignore_findings":[],"done":true,"input_fingerprint":"791748764cdc0dfe81d987a908179930d858670606e51b64407172295776b261","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Prepare the comparison"},"impact":2,"gaps":0,"r":48,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","label":"Let an agent inspect the e...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","acceptance":["The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","get_evidence exposes the selected finding's numerical support and relationship source lineage.","Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated."],"guidance":["Run results are retained only in the bounded server-session cache; missing or evicted run IDs require reanalysis.","Cases, episodes, and residual patterns may overlap."],"path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":["interfaces/mcp/connect-read-only-local-server","interfaces/skill/guide-anomaly-investigation"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py"],"automatedChecks":[{"id":"case-drilldown","description":"Analyze, page findings, and inspect exact conversion case values and lineage.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case tests/test_mcp.py::test_schema_11_episode_is_paged_and_has_relationship_evidence"},{"id":"unavailable-finding","description":"Expose a quality finding when latest relationship evidence is unavailable.","command":"python -m pytest -q tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"},{"id":"early-candidate","description":"Inspect descriptive candidate evidence without inventing a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case tests/test_mcp.py::test_sparse_insufficient_history_keeps_descriptive_candidate_visible"},{"id":"sample-pagination","description":"Page supporting samples for a residual pattern.","command":"python -m pytest -q tests/test_mcp.py::test_get_evidence_pages_pattern_samples"}],"manualChecks":["Follow analyze, list_findings, and get_evidence for the conversion example and check that the headline is supported by the retrieved values."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","criteria":[{"criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"covered","evidence":["src/anomalyzer/mcp.py:170","src/anomalyzer/mcp_views.py:247","tests/test_mcp.py:54"],"notes":"The conversion manual call returns a relationship headline, preview, counts, readiness, and session-scoped run ID."},{"criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:132","src/anomalyzer/mcp_views.py:260","tests/test_mcp.py:313","tests/test_mcp.py:108"],"notes":"Cases, early candidates, episodes, patterns, and quality findings feed a bounded page; relationship episode regression verifies schema 1.1 pagination."},{"criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:273","src/anomalyzer/mcp_views.py:294","tests/test_mcp.py:54","tests/test_mcp.py:313"],"notes":"Manual drilldown returns observed 8/130 versus expected 0.1 and source indexes orders=14 and qualified_visits=14."},{"criterion":"Unavailable latest evidence and non-triggering early candidates remain visible, while large pattern sample sets are paginated.","status":"covered","evidence":["src/anomalyzer/mcp_views.py:118","src/anomalyzer/mcp_views.py:169","src/anomalyzer/mcp_views.py:337","tests/test_mcp.py:87","tests/test_mcp.py:137","tests/test_mcp.py:341"],"notes":"Passing tests retain short-history and inapplicable-method candidates plus unavailable latest evidence, while pattern samples are paginated."}],"ignore_findings":[],"done":true,"input_fingerprint":"a09b0003780bcd2f9ccf8e3d8ff58b3147c118760b984757d7027b5810fd9bef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Drill into findings and evidence"},"impact":2,"gaps":0,"r":48,"x":927.3333333333334,"y":288},{"id":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","label":"Spot early changes with a ...","area":"investigation","scopes":["investigation","investigation/assessments"],"scope":"investigation/assessments","scopeLabel":"investigation/assessments","status":"implemented","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","acceptance":["The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","Missing values break the comparison history, and zero previous values do not create relative-change evidence.","Early candidates remain non-triggering and ineligible for action even under a permissive policy."],"guidance":["An early candidate does not establish an anomaly, stable baseline, cause, or incident."],"path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":["investigation/cases/group-related-evidence","investigation/rules/check-declared-numerical-rules"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_mcp.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"prefix-invariant-facts","description":"Describe short history causally without creating cases.","command":"python -m pytest -q tests/test_agentic.py::test_short_history_assessments_are_useful_and_prefix_invariant"},{"id":"comparison-boundaries","description":"Reset comparisons at missing observations and avoid dividing by zero.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"candidate-eligibility","description":"Keep candidates ineligible for action and notifications under permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"},{"id":"agent-visibility","description":"Expose a short-history candidate and its claim limits without a case.","command":"python -m pytest -q tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case"}],"manualChecks":["Inspect a candidate's comparison population, maturity, established claim, and not_established text before describing it."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","criteria":[{"criterion":"The first observation records a value, the second records a change, and later departures from the prior range can be candidates.","status":"covered","evidence":["src/anomalyzer/early.py:73","src/anomalyzer/early.py:96","tests/test_agentic.py:35"],"notes":"Assertions establish observation/change/change/departure_candidate for [10,12,11,20]."},{"criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"covered","evidence":["src/anomalyzer/early.py:50","src/anomalyzer/early.py:141","tests/test_agentic.py:35"],"notes":"Each assessment precedes the prior-history update; tests compare complete serialized assessments for every prefix."},{"criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"covered","evidence":["src/anomalyzer/early.py:53","src/anomalyzer/early.py:58","tests/test_capability_behaviors.py:87"],"notes":"The zero comparison has no relative change; after a missing value, comparison samples reset to zero."},{"criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"covered","evidence":["src/anomalyzer/early.py:105","src/anomalyzer/cases.py:35","tests/test_capability_behaviors.py:134","tests/test_mcp.py:341"],"notes":"Candidates have criterion_met=false and remain action/notification-ineligible after policy replay, without creating cases."}],"ignore_findings":[],"done":true,"input_fingerprint":"debdeeeaea559950f53461a4298cb9c2e3a1681f83a43e8a981ed9b51968a4d0","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect early evidence"},"impact":2,"gaps":0,"r":48,"x":1110.6666666666667,"y":288},{"id":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","label":"Compare the same entity an...","area":"relationships","scopes":["relationships","relationships/quality"],"scope":"relationships/quality","scopeLabel":"relationships/quality","status":"implemented","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","acceptance":["Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","Difference relationships require identical declared source units and matching output units.","An invalid relationship is reported as failed while valid sibling dataset results remain available."],"guidance":["A failed or unavailable relationship does not establish that its metric is normal.","No implicit timestamp matching, interpolation, unit conversion, or zero-filling occurs."],"path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"implementationReferences":["src/anomalyzer/alignment.py","src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"exact-alignment","description":"Reject mismatched timestamps, cadence, entity, duplicates, and positional relationship sources.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relationship_requires_exact_source_alignment"},{"id":"compatible-units","description":"Report a difference unit mismatch without suppressing sibling results.","command":"python -m pytest -q tests/test_agentic.py::test_alignment_unit_and_partial_failure_are_specific"},{"id":"unambiguous-declaration","description":"Reject fields that do not belong to the selected relationship kind.","command":"python -m pytest -q tests/test_agentic.py::test_contract_rejects_ambiguous_relationship_fields"},{"id":"healthy-siblings","description":"Retain a healthy dataset's evidence when another dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Read both target-level statuses and errors before interpreting a partial run; check source dataset errors when a relationship source failed."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","criteria":[{"criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"covered","evidence":["src/anomalyzer/alignment.py:42","tests/test_capability_behaviors.py:30"],"notes":"Parameterized assertions require specific errors for timestamp, cadence, and entity mismatches."},{"criterion":"Relationship sources are timestamped and have unique timestamps; implicit resampling and duplicate aggregation are not accepted for lineage.","status":"covered","evidence":["src/anomalyzer/alignment.py:23","src/anomalyzer/alignment.py:44","tests/test_capability_behaviors.py:30"],"notes":"Positional sources fail exact alignment; even an explicit duplicate-mean setting is rejected for relationship source lineage."},{"criterion":"Difference relationships require identical declared source units and matching output units.","status":"covered","evidence":["src/anomalyzer/relationships.py:41","tests/test_agentic.py:242"],"notes":"Source and output unit checks are explicit. Tests cover mismatched source units and successful matching units."},{"criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:171","tests/test_capability_behaviors.py:30","tests/test_capability_behaviors.py:159"],"notes":"Relationship errors remain target-specific, healthy dataset results survive, and the overall mixed run is partial."}],"ignore_findings":[],"done":true,"input_fingerprint":"06321ab5806a8276cc3954df5d6f4dd7f059f67600038854569787ff51b420d6","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Align comparable sources"},"impact":2,"gaps":0,"r":48,"x":1294,"y":288},{"id":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","label":"Compare against a new base...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","acceptance":["Integer positions and exact observed dates or timestamps can identify reset boundaries.","Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","Repeated, out-of-range, or unmatched reset points are rejected.","CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries."],"guidance":["Reset points are explicit caller decisions. Known-event notes and review-state fields do not rebuild regimes.","Multi-resolution analysis does not support reset_points."],"path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["data/quality/analyze-sparse-observations"],"implementationReferences":["src/anomalyzer/validation.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py"],"automatedChecks":[{"id":"independent-regimes","description":"Retrain separate regimes without cross-boundary evidence or patterns.","command":"python -m pytest -q tests/test_analysis.py::test_manual_resets_retrain_without_crossing_boundaries"},{"id":"invalid-boundaries","description":"Reject invalid or duplicate resolved reset positions.","command":"python -m pytest -q tests/test_analysis.py::test_invalid_manual_resets"},{"id":"cli-boundaries","description":"Resolve position and date reset flags through the CLI.","command":"python -m pytest -q tests/test_cli.py::test_manual_reset_cli"}],"manualChecks":["Inspect the resolved reset positions and segment diagnostics for examples/steps_reset.json."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","criteria":[{"criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"covered","evidence":["src/anomalyzer/validation.py:129","tests/test_analysis.py:381"],"notes":"Positions and observed dates are exercised end to end; timestamp normalization and exact matching are implemented directly."},{"criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"covered","evidence":["src/anomalyzer/recipes.py:119","src/anomalyzer/recipes.py:143","tests/test_analysis.py:381"],"notes":"Each segment initializes separate state and passes its own start to forecasts. Tests verify three calibrated regimes and no cross-boundary evidence or findings."},{"criterion":"Repeated, out-of-range, or unmatched reset points are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/validation.py:149","tests/test_analysis.py:413"],"notes":"Validation rejects literal duplicates and different declarations resolving to the same position, invalid positions, and unmatched dates/timestamps."},{"criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"covered","evidence":["tests/test_cli.py:82","src/anomalyzer/recipes.py:53"],"notes":"CLI assertions establish positions [50, 100] and removal of the unmarked-regime episodes."}],"ignore_findings":[],"done":true,"input_fingerprint":"067063f972c5b70d990d6cb0125892731a97a7f482698d7b331a1a30229ac223","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Restart after a reviewed change"},"impact":1,"gaps":0,"r":43,"x":194,"y":288},{"id":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","label":"Assess sparse observations...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Schema-1.1 callers declare coverage and missing-data semantics per source, keeping unknown collection distinct from confirmed empty additive volume.","intent":"Retain useful rule and supported statistical evidence for sparse histories while preserving time, provenance, incomplete periods and evidence limits.","acceptance":["Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals."],"guidance":["Default missingness is unknown. A modeled expectation is never a measurement.","Caller business severity, incident state and notifications remain outside the engine."],"path":".capabilities/data/quality/analyze-sparse-observations.capability.yaml","dependencies":["data/inputs/accept-analysis-requests","analysis/expectations/reset-declared-regimes","analysis/detection/identify-residual-patterns"],"suggestedDependencies":[],"dependents":["relationships/resolution/compare-finalized-subperiods"],"implementationReferences":["src/anomalyzer/contracts.py","src/anomalyzer/sparse.py","src/anomalyzer/sparse_views.py","src/anomalyzer/orchestrator.py","src/anomalyzer/relationships.py","src/anomalyzer/early.py","tests/test_sparse.py","docs/sparse-data.md","examples/sparse_examples.py","examples/render_readme_progression.py"],"automatedChecks":[{"id":"source-semantics","description":"Preserve unknown periods and require explicit empty-volume authorization.","command":"python -m pytest -q tests/test_sparse.py -k \"zero or omitted or grid or duplicate or expansion\""},{"id":"causal-boundaries","description":"Retain real phase, event times, resets and as-of revision eligibility.","command":"python -m pytest -q tests/test_sparse.py -k \"phase or gap or reset or event or revision or arrival\""},{"id":"public-interfaces","description":"Preserve useful rules, relationships, views, cases and API parity.","command":"python -m pytest -q tests/test_sparse.py -k \"rule or relationship or daily or parity\""},{"id":"progressive-examples","description":"Reproduce the full numbered sequence and sparse worked examples.","command":"python examples/run_demo.py"}],"manualChecks":["Inspect progression figures 14–22 for true gaps, distinguished policy zeros, original event time, unavailable checks and caller-owned business severity."],"verificationGaps":[],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for data/quality/analyze-sparse-observations.","criteria":[{"criterion":"Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","status":"partial","evidence":["src/anomalyzer/contracts.py:51","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:71","src/anomalyzer/contracts.py:85","src/anomalyzer/contracts.py:86","src/anomalyzer/sparse.py:3","src/anomalyzer/sparse.py:34","src/anomalyzer/sparse.py:42","src/anomalyzer/sparse.py:48","src/anomalyzer/sparse.py:54","src/anomalyzer/sparse_views.py:74","src/anomalyzer/sparse_views.py:81","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:97","src/anomalyzer/sparse_views.py:119","src/anomalyzer/orchestrator.py:63","src/anomalyzer/orchestrator.py:127","src/anomalyzer/orchestrator.py:139","src/anomalyzer/orchestrator.py:191","src/anomalyzer/relationships.py:36","src/anomalyzer/relationships.py:58","src/anomalyzer/relationships.py:98","src/anomalyzer/relationships.py:131","src/anomalyzer/relationships.py:141","src/anomalyzer/early.py:58","src/anomalyzer/early.py:137","tests/test_sparse.py:44","tests/test_sparse.py:45","tests/test_sparse.py:50","tests/test_sparse.py:51","tests/test_sparse.py:57","docs/sparse-data.md:4","docs/sparse-data.md:29","docs/sparse-data.md:37","docs/sparse-data.md:39","docs/sparse-data.md:40","examples/sparse_examples.py:22","examples/sparse_examples.py:23","examples/sparse_examples.py:25","examples/sparse_examples.py:26","examples/sparse_examples.py:28","examples/render_readme_progression.py:135","examples/render_readme_progression.py:139","examples/render_readme_progression.py:140","examples/render_readme_progression.py:166","examples/render_readme_progression.py:206"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","status":"partial","evidence":["src/anomalyzer/contracts.py:27","src/anomalyzer/contracts.py:49","src/anomalyzer/contracts.py:51","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:56","src/anomalyzer/sparse.py:1","src/anomalyzer/sparse.py:23","src/anomalyzer/sparse.py:181","src/anomalyzer/sparse.py:187","src/anomalyzer/sparse.py:189","src/anomalyzer/sparse_views.py:15","src/anomalyzer/sparse_views.py:98","src/anomalyzer/sparse_views.py:107","src/anomalyzer/sparse_views.py:108","src/anomalyzer/sparse_views.py:112","src/anomalyzer/orchestrator.py:22","src/anomalyzer/orchestrator.py:23","src/anomalyzer/orchestrator.py:38","src/anomalyzer/orchestrator.py:129","src/anomalyzer/orchestrator.py:141","src/anomalyzer/relationships.py:185","src/anomalyzer/relationships.py:189","src/anomalyzer/relationships.py:191","src/anomalyzer/early.py:46","src/anomalyzer/early.py:48","src/anomalyzer/early.py:51","src/anomalyzer/early.py:59","src/anomalyzer/early.py:61","tests/test_sparse.py:20","tests/test_sparse.py:108","tests/test_sparse.py:111","tests/test_sparse.py:113","tests/test_sparse.py:119","docs/sparse-data.md:37","docs/sparse-data.md:41","docs/sparse-data.md:42","docs/sparse-data.md:106","docs/sparse-data.md:108","examples/sparse_examples.py:16","examples/sparse_examples.py:33","examples/sparse_examples.py:50","examples/sparse_examples.py:51","examples/sparse_examples.py:91","examples/render_readme_progression.py:241","examples/render_readme_progression.py:312","examples/render_readme_progression.py:317","examples/render_readme_progression.py:346","examples/render_readme_progression.py:349"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","status":"partial","evidence":["src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:81","src/anomalyzer/contracts.py:242","src/anomalyzer/contracts.py:245","src/anomalyzer/contracts.py:258","src/anomalyzer/sparse.py:4","src/anomalyzer/sparse.py:49","src/anomalyzer/sparse.py:57","src/anomalyzer/sparse.py:63","src/anomalyzer/sparse.py:64","src/anomalyzer/sparse_views.py:75","src/anomalyzer/sparse_views.py:94","src/anomalyzer/sparse_views.py:127","src/anomalyzer/sparse_views.py:155","src/anomalyzer/sparse_views.py:157","src/anomalyzer/orchestrator.py:37","src/anomalyzer/orchestrator.py:38","src/anomalyzer/orchestrator.py:166","src/anomalyzer/orchestrator.py:178","src/anomalyzer/orchestrator.py:180","src/anomalyzer/relationships.py:40","src/anomalyzer/relationships.py:57","src/anomalyzer/relationships.py:58","src/anomalyzer/relationships.py:96","src/anomalyzer/relationships.py:97","src/anomalyzer/early.py:140","src/anomalyzer/early.py:147","tests/test_sparse.py:59","tests/test_sparse.py:67","tests/test_sparse.py:75","tests/test_sparse.py:99","tests/test_sparse.py:128","docs/sparse-data.md:26","docs/sparse-data.md:96","docs/sparse-data.md:101","docs/sparse-data.md:132","docs/sparse-data.md:142","examples/sparse_examples.py:26","examples/sparse_examples.py:36","examples/sparse_examples.py:38","examples/sparse_examples.py:57","examples/sparse_examples.py:81","examples/render_readme_progression.py:654","examples/render_readme_progression.py:697","examples/render_readme_progression.py:884","examples/render_readme_progression.py:896","examples/render_readme_progression.py:917"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals.","status":"partial","evidence":["src/anomalyzer/contracts.py:13","src/anomalyzer/contracts.py:54","src/anomalyzer/contracts.py:297","src/anomalyzer/contracts.py:298","src/anomalyzer/contracts.py:437","src/anomalyzer/sparse.py:4","src/anomalyzer/sparse.py:34","src/anomalyzer/sparse.py:64","src/anomalyzer/sparse.py:66","src/anomalyzer/sparse.py:83","src/anomalyzer/sparse_views.py:1","src/anomalyzer/sparse_views.py:28","src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:94","src/anomalyzer/orchestrator.py:106","src/anomalyzer/orchestrator.py:107","src/anomalyzer/orchestrator.py:133","src/anomalyzer/orchestrator.py:166","src/anomalyzer/orchestrator.py:180","src/anomalyzer/relationships.py:160","src/anomalyzer/relationships.py:162","src/anomalyzer/relationships.py:166","tests/test_sparse.py:59","tests/test_sparse.py:75","tests/test_sparse.py:99","tests/test_sparse.py:123","tests/test_sparse.py:180","docs/sparse-data.md:26","docs/sparse-data.md:41","docs/sparse-data.md:69","docs/sparse-data.md:78","docs/sparse-data.md:96","examples/sparse_examples.py:59","examples/sparse_examples.py:95","examples/render_readme_progression.py:686","examples/render_readme_progression.py:699","examples/render_readme_progression.py:710","examples/render_readme_progression.py:767","examples/render_readme_progression.py:884"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","partial: Frozen eligible training and calibration support point references across gaps, while pattern and change comparisons respect gaps and explicit resets.","partial: Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","partial: As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals."],"reviewStale":false,"health":"gap","healthLabel":"Review findings","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Assess sparse source data","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Assess sparse source data"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","label":"Avoid mistaking unfinished...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"implemented","summary":"Explicitly incomplete observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","acceptance":["Schema-1.0 incomplete observations form a trailing suffix; schema-1.1 source policies preserve internal and trailing incomplete positions as unavailable.","The observed values and timestamps of unfinished periods remain available in data quality output.","Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","A relationship with an unavailable latest source reports that unavailability rather than normal behavior."],"guidance":["Completeness is declared by the caller; the engine does not infer collection completeness."],"path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","dependencies":["data/quality/prepare-timestamped-series"],"suggestedDependencies":[],"dependents":["analysis/resolution/monitor-within-unfinished-days"],"implementationReferences":["src/anomalyzer/sparse.py","src/anomalyzer/sparse_views.py","tests/test_sparse.py","src/anomalyzer/validation.py","src/anomalyzer/alignment.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_mcp.py"],"automatedChecks":[{"id":"unfinished-values","description":"Retain an unfinished value without scoring it.","command":"python -m pytest -q tests/test_analysis.py::test_explicit_incomplete_period_is_reported_and_not_scored"},{"id":"trailing-only","description":"Reject incomplete observations followed by completed observations.","command":"python -m pytest -q tests/test_analysis.py::test_incomplete_periods_must_be_trailing"},{"id":"source-and-agent-views","description":"Exclude a partial source sample from both analysis views and expose unavailable relationship evidence.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days tests/test_mcp.py::test_agent_sees_unavailable_latest_evidence"}],"manualChecks":["Inspect examples/agentic/incomplete-usage.json and confirm the latest headline explains missing completeness and does not imply normality."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","criteria":[{"criterion":"Incomplete observations form a trailing suffix after chronological preparation.","status":"covered","evidence":["src/anomalyzer/validation.py:44","tests/test_analysis.py:162"],"notes":"Suffix validation runs after timestamp preparation and rejects a completed observation following an incomplete one."},{"criterion":"The observed values and timestamps of unfinished periods remain available in data quality output.","status":"covered","evidence":["src/anomalyzer/validation.py:50","src/anomalyzer/validation.py:114","tests/test_analysis.py:150"],"notes":"Quality records retain original source index, normalized timestamp, observed value, and incomplete status."},{"criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"covered","evidence":["src/anomalyzer/validation.py:53","src/anomalyzer/recipes.py:564","tests/test_analysis.py:150","tests/test_analysis.py:200"],"notes":"Only finalized prepared values enter either scoring path; assertions verify the last scored sample and both multi-resolution counts."},{"criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"covered","evidence":["src/anomalyzer/alignment.py:35","src/anomalyzer/orchestrator.py:150","tests/test_mcp.py:87"],"notes":"Incomplete aligned sources become unavailable values, and latest relationship state/status plus MCP headline explain the lack of evidence."}],"ignore_findings":[],"done":true,"input_fingerprint":"d82ee816fe89a25232e2589587d31bcadfa81be4188a0e1531f4fbfdf0ebcb5e","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Keep unfinished data visible"},"impact":1,"gaps":0,"r":43,"x":377.33333333333337,"y":288},{"id":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","label":"Give an agent read-only ac...","area":"interfaces","scopes":["interfaces","interfaces/mcp"],"scope":"interfaces/mcp","scopeLabel":"interfaces/mcp","status":"implemented","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","acceptance":["Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","Analysis tool calls return structured content together with compact text.","Invalid tool input and unknown session run IDs return structured errors rather than fabricated results."],"guidance":["Compatibility tool names remain callable but are not advertised.","Legacy clients receive isError tool results; modern clients receive JSON-RPC invalid-input errors.","The server provides analysis and evidence inspection; the caller owns monitoring and operational workflows."],"path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","dependencies":["interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":["interfaces/python/install-versioned-library"],"implementationReferences":["src/anomalyzer/mcp.py","src/anomalyzer/mcp_views.py","tests/test_mcp.py","pyproject.toml"],"automatedChecks":[{"id":"tool-discovery","description":"Advertise the three read-only tools and their schema contracts.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_lists_read_only_tools"},{"id":"connection-protocols","description":"Support legacy stdio initialization and modern stateless discovery.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_stdio_legacy_handshake_and_tool_listing tests/test_mcp.py::test_mcp_modern_discovery_and_stateless_tool_listing"},{"id":"structured-content","description":"Return a compact analysis response with structured content and run ID.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_tool_call_returns_structured_content"},{"id":"invalid-calls","description":"Reject malformed input and unknown session run IDs in both protocols.","command":"python -m pytest -q tests/test_mcp.py::test_mcp_invalid_tool_input_is_a_tool_error tests/test_mcp.py::test_paged_tool_inputs_enforce_schema_in_both_protocols"}],"manualChecks":["Review discovery's interpretation guidance and confirm the client can start anomalyzer-mcp in the installed Python environment."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","criteria":[{"criterion":"Tool discovery advertises analyze, list_findings, and get_evidence as read-only with input and output schemas.","status":"covered","evidence":["src/anomalyzer/mcp.py:63","tests/test_mcp.py:27"],"notes":"Discovery exposes three read-only tools with the same strict argument contracts used at dispatch and advertised output schemas."},{"criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"covered","evidence":["src/anomalyzer/mcp.py:228","tests/test_mcp.py:248","tests/test_mcp.py:268"],"notes":"Legacy initialization and modern stateless discovery tests establish protocol metadata and evidence interpretation guidance."},{"criterion":"Analysis tool calls return structured content together with compact text.","status":"covered","evidence":["src/anomalyzer/mcp.py:162","tests/test_mcp.py:39"],"notes":"The manual conversion call returns a compact headline and structured run ID, counts, readiness, and preview."},{"criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"covered","evidence":["src/anomalyzer/mcp.py:156","src/anomalyzer/mcp.py:257","tests/test_mcp.py:235","tests/test_mcp.py:367"],"notes":"Tests reject invalid cursor/limit types, unknown fields, numeric run IDs, and missing session runs in both protocols without fabricated results."}],"ignore_findings":[],"done":true,"input_fingerprint":"7bd92d238f4f9545372c6c35e78aefea60c585bea0a4a8bb99096ab6cef27f96","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","name":"Read-only MCP server","label":"Read-only MCP server / Give an agent access to local evidence / Connect a local server"},"impact":1,"gaps":0,"r":43,"x":903.3333333333334,"y":406},{"id":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","label":"Find values that break my ...","area":"investigation","scopes":["investigation","investigation/rules"],"scope":"investigation/rules","scopeLabel":"investigation/rules","status":"implemented","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","acceptance":["Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","Absolute and relative changes must exceed their declared maximum to violate a rule.","Relative-change checks require a nonzero previous value and do not bridge missing observations.","Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them."],"guidance":["An exact rule crossing does not establish cause, an incident, or business impact."],"path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","dependencies":["investigation/assessments/describe-early-changes"],"suggestedDependencies":[],"dependents":["investigation/policy/replay-action-eligibility"],"implementationReferences":["src/anomalyzer/early.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"rule-boundaries","description":"Check inclusive range boundaries and strict absolute/relative change thresholds.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_explicit_rules_respect_boundaries"},{"id":"relative-availability","description":"Avoid relative violations across zero or missing values.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_relative_rule_does_not_compare_across_zero_or_missing_values"},{"id":"explicit-not-actionable","description":"Create evidence and a case for an exact rule crossing without enabling action by default.","command":"python -m pytest -q tests/test_agentic.py::test_explicit_rule_is_exact_but_not_actionable_without_policy"},{"id":"readable-rule","description":"Identify the exact unit-cost range rule in CLI output.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the unit-cost example's crossed rule and confirm its text distinguishes the declared bound from statistical detection."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","criteria":[{"criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"covered","evidence":["src/anomalyzer/early.py:22","src/anomalyzer/contracts.py:209","tests/test_capability_behaviors.py:74"],"notes":"Boundary values 10 and 20 are accepted; 9 and 21 violate the declared range."},{"criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"covered","evidence":["src/anomalyzer/early.py:25","tests/test_capability_behaviors.py:74"],"notes":"Parameterized tests distinguish equality to the threshold from strict exceedance for both change rules."},{"criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"covered","evidence":["src/anomalyzer/early.py:29","src/anomalyzer/early.py:53","tests/test_capability_behaviors.py:87"],"notes":"Zero/missing predecessors suppress relative comparisons, and missing values reset comparison history."},{"criterion":"Violations identify the exact declared rule and remain action-ineligible unless policy explicitly permits them.","status":"covered","evidence":["src/anomalyzer/early.py:34","src/anomalyzer/early.py:121","tests/test_agentic.py:49","tests/test_cli.py:134"],"notes":"Violation records retain rule ID/kind/bounds; default eligibility is false and replay requires explicit permission."}],"ignore_findings":[],"done":true,"input_fingerprint":"60d24e84faee9c1e2ae8f9337d17334b37c89305093404da2e84c51526111d0b","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Apply my stated limits"},"impact":1,"gaps":0,"r":43,"x":1086.6666666666667,"y":406},{"id":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","label":"Find changes in conversion...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","acceptance":["Ratios retain the computed value, formula, and original source indexes for each sample.","Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","Derived series can be analyzed with the frozen or adaptive single-resolution recipe."],"guidance":["Relationships are caller-supplied declarations, not discovered or causal relationships.","A difference requires identical declared units; a normalized residual requires a positive supplied scale.","A joint condition is a declared 0/1 comparison, not statistical proof."],"path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","dependencies":["relationships/quality/require-compatible-sources","analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":["relationships/resolution/compare-finalized-subperiods"],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_examples.py"],"automatedChecks":[{"id":"ratio-lineage","description":"Explain a changing rate with the exact source indexes and formula.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"other-comparisons","description":"Check declared differences, normalized residuals, and combined conditions.","command":"python -m pytest -q tests/test_agentic.py::test_remaining_relationship_primitives"},{"id":"unavailable-values","description":"Keep zero-denominator and missing-source observations unavailable.","command":"python -m pytest -q tests/test_agentic.py::test_missing_and_zero_denominator_are_unavailable_not_zero"},{"id":"adaptive-relationship","description":"Use the adaptive recipe for a declared relationship.","command":"python -m pytest -q tests/test_agentic.py::test_adaptive_recipe_runs_for_declared_relationships"}],"manualChecks":["Inspect examples/agentic/conversion.json and distinguish its relationship departure from the individual source metrics."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","criteria":[{"criterion":"Ratios retain the computed value, formula, and original source indexes for each sample.","status":"covered","evidence":["src/anomalyzer/relationships.py:84","src/anomalyzer/relationships.py:152","tests/test_agentic.py:211"],"notes":"Assertions check exact ratio values, formula, and both original source indexes."},{"criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"covered","evidence":["src/anomalyzer/relationships.py:97","src/anomalyzer/relationships.py:134","tests/test_agentic.py:285"],"notes":"Parameterized assertions independently specify difference, normalized residual, all-condition, and any-condition outputs."},{"criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"covered","evidence":["src/anomalyzer/relationships.py:89","src/anomalyzer/relationships.py:142","tests/test_agentic.py:226"],"notes":"Lineage retains null values with missing_source_value/zero_denominator reasons. Explicit zero_denominator=error intentionally fails instead, as documented."},{"criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:87","src/anomalyzer/orchestrator.py:135","tests/test_agentic.py:171","tests/test_examples.py:17"],"notes":"Both supported recipes dispatch derived datasets through the shared single-resolution engine."}],"ignore_findings":[],"done":true,"input_fingerprint":"7fa8f7ffdc0d852ad961b3e1753761d1bd3b00b87b8c001a957ec19e26d068b9","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Compare ratios and metric gaps"},"impact":1,"gaps":0,"r":43,"x":1270,"y":406},{"id":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","label":"Find unusual spikes and dr...","area":"analysis","scopes":["analysis","analysis/detection"],"scope":"analysis/detection","scopeLabel":"analysis/detection","status":"implemented","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","acceptance":["A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","Separated point departures remain separate episodes, and positional episode intervals use integer coordinates."],"guidance":["Detection is a numerical departure from the selected expectation, not a claim about cause or business impact."],"path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/recipes.py","src/anomalyzer/evidence.py","tests/test_analysis.py"],"automatedChecks":[{"id":"point-evidence","description":"Detect an injected spike and retain deterministic earlier numerical evidence.","command":"python -m pytest -q tests/test_analysis.py::test_spike_and_no_lookahead"},{"id":"adjacent-run","description":"Describe adjacent point triggers and their contributing positions.","command":"python -m pytest -q tests/test_analysis.py::test_anomaly_stream_detects_consecutive_run"},{"id":"isolated-events","description":"Keep separated anomalies from becoming a consecutive-run pattern.","command":"python -m pytest -q tests/test_analysis.py::test_isolated_anomalies_do_not_form_second_order_pattern"},{"id":"positional-episodes","description":"Use integer intervals for position-only episode evidence.","command":"python -m pytest -q tests/test_analysis.py::test_positional_episode_uses_integer_interval"}],"manualChecks":["Inspect observations and triggering_samples for examples/spike.json and distinguish episode grouping from pattern findings."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","criteria":[{"criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"covered","evidence":["src/anomalyzer/recipes.py:344","src/anomalyzer/recipes.py:388","tests/test_analysis.py:441"],"notes":"The injected spike retains its observed value, expectation, standardized residual, and point trigger."},{"criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"covered","evidence":["src/anomalyzer/baseline.py:31","src/anomalyzer/recipes.py:174","tests/test_analysis.py:441","tests/test_trend.py:36"],"notes":"Assertions establish repeatability and unchanged earlier evidence under later evaluation changes with fixed seasonal settings. Automatic positional season discovery has the separately documented streaming limitation."},{"criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"covered","evidence":["src/anomalyzer/evidence.py:95","tests/test_analysis.py:274"],"notes":"The test checks one three-sample run, exact contributing positions, direction, and first detection."},{"criterion":"Separated point departures remain separate episodes, and positional episode intervals use integer coordinates.","status":"covered","evidence":["src/anomalyzer/evidence.py:12","src/anomalyzer/evidence.py:23","tests/test_analysis.py:263","tests/test_analysis.py:369"],"notes":"Separated spikes produce two episodes without a consecutive-run pattern; positional evidence produces interval [9, 9]."}],"ignore_findings":[],"done":true,"input_fingerprint":"062833cc1bd19d5734b0f3c63d1e81666ffb568ead3399e359d62db2c0c0677d","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Find spikes and drops"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","label":"Know when the expected bas...","area":"analysis","scopes":["analysis","analysis/evidence"],"scope":"analysis/evidence","scopeLabel":"analysis/evidence","status":"implemented","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","acceptance":["Repeating seasonal structure with stable calibration can yield supported readiness.","Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","Insufficient calibration support yields not_assessed, including when the run otherwise completes.","Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment."],"guidance":["Readiness does not prove real-world effectiveness and is not an automatic action-policy gate."],"path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/readiness.py","src/anomalyzer/recipes.py","tests/test_readiness.py"],"automatedChecks":[{"id":"supported-reference","description":"Support repeated seasonality and preserve readiness under later changes.","command":"python -m pytest -q tests/test_readiness.py::test_repeating_season_and_trend_support_reference"},{"id":"weak-season","description":"Warn on unstable seasonality even when no findings occur.","command":"python -m pytest -q tests/test_readiness.py::test_unstable_declared_season_warns_without_calling_series_inapplicable tests/test_readiness.py::test_documented_random_example_has_no_findings_but_warns_on_reference"},{"id":"calibration-drift","description":"Warn when residual calibration drifts despite a repeated training season.","command":"python -m pytest -q tests/test_readiness.py::test_calibration_drift_warns_even_when_training_season_repeats"},{"id":"not-assessed","description":"Distinguish uncalibrated history and too few calibration samples.","command":"python -m pytest -q tests/test_readiness.py::test_readiness_is_not_assessed_without_calibrated_evaluation tests/test_readiness.py::test_short_calibration_is_not_assessed_even_when_analysis_completes"}],"manualChecks":["Read readiness reasons alongside findings; a lack of findings alone does not establish a trustworthy reference."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","criteria":[{"criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"covered","evidence":["src/anomalyzer/readiness.py:128","tests/test_readiness.py:8"],"notes":"Assertions establish repeated seasonality, strong calibration improvement, and supported readiness."},{"criterion":"Unstable declared seasonality or a drifting calibration center yields caution without making an otherwise completed run fail.","status":"covered","evidence":["src/anomalyzer/readiness.py:113","src/anomalyzer/readiness.py:122","tests/test_readiness.py:28","tests/test_readiness.py:41"],"notes":"Completed analyses retain caution reasons for weak seasonal repetition or calibration-center drift."},{"criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"covered","evidence":["src/anomalyzer/readiness.py:77","tests/test_readiness.py:56","tests/test_readiness.py:63"],"notes":"Both uncalibrated history and a completed run with three calibration samples return not_assessed."},{"criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"covered","evidence":["src/anomalyzer/recipes.py:327","tests/test_readiness.py:8"],"notes":"Readiness reads training/calibration inputs once; tests compare identical segment and aggregate assessments after a later shift."}],"ignore_findings":[],"done":true,"input_fingerprint":"d01367627e058ea39c78ee0e3616061a2e0725d06e8acd0b84b46a9655431e57","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Assess baseline quality"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","label":"Follow evolving growth whi...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","acceptance":["Each rolling expectation uses only observations before the sample being scored.","Persistent growth changes can be followed more closely than by the frozen default recipe.","An isolated extreme value is held out of future model references rather than creating a seasonal echo.","Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected."],"guidance":["A persistent third extreme miss in one direction may be admitted to the rolling reference; this is not a manual regime reset."],"path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","dependencies":["analysis/evidence/calibrate-departures"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/baseline.py","src/anomalyzer/recipes.py","src/anomalyzer/contracts.py","tests/test_analysis.py"],"automatedChecks":[{"id":"causal-window","description":"Keep a known rolling expectation invariant when future values are appended.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_expectation_is_one_step_and_prefix_invariant"},{"id":"changing-growth","description":"Follow a persistent growth change and report model admissions.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_tracks_a_persistent_growth_change"},{"id":"isolated-extreme","description":"Exclude an isolated extreme from later reference estimates.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_recipe_contains_an_isolated_extreme"},{"id":"bounded-settings","description":"Reject conflicting or unbounded adaptive settings.","command":"python -m pytest -q tests/test_analysis.py::test_adaptive_settings_reject_conflicting_or_unbounded_trend"}],"manualChecks":["Inspect the adaptive recipe's rolling-window parameters and frozen calibration before interpreting its final expectation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","criteria":[{"criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"covered","evidence":["src/anomalyzer/baseline.py:61","tests/test_analysis.py:48"],"notes":"All rolling-reference positions precede the scored index. The independent expectation remains unchanged after future values are appended."},{"criterion":"Persistent growth changes can be followed more closely than by the frozen default recipe.","status":"covered","evidence":["src/anomalyzer/recipes.py:145","src/anomalyzer/recipes.py:370","tests/test_analysis.py:58"],"notes":"The test requires adaptive held-out MAE below half the frozen recipe's MAE and a close final expectation."},{"criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","tests/test_analysis.py:81"],"notes":"The isolated spike is excluded from model references; its next seasonal expectation closely matches the control. Persistent third misses are admitted as guidance explicitly permits."},{"criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"covered","evidence":["src/anomalyzer/contracts.py:41","tests/test_analysis.py:104"],"notes":"Both incompatible explicit trend selection and excessive slope lookback raise specific validation errors."}],"ignore_findings":[],"done":true,"input_fingerprint":"984422f1fd3ab3d22cf7b4ba9d0bd06629fc228a8e53d7ad44439cd8d35ec0db","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Follow evolving growth"},"impact":0,"gaps":0,"r":38,"x":170,"y":406},{"id":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","label":"Account for daily and week...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"planned","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","acceptance":["Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","Insufficient support for a requested cycle produces an explicit limitation or declared fallback rather than a silently fitted component.","Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle."],"guidance":["Begin with a bounded daily-and-weekly use case; overlapping cycles share information and cannot be treated as independent evidence.","Proposed automated checks cover combined ordinary cycles, future isolation, unsupported cycle history, and extreme-value influence across cycles.","Manual review should inspect a weekday morning spike and verify the baseline explains ordinary calendar variation without explaining away the spike."],"path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Represent more than one repeating cycle"},"impact":0,"gaps":3,"r":47,"x":170,"y":406},{"id":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","label":"Limit how extreme values d...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","acceptance":["Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","Training and calibration diagnostics identify excluded extreme observations.","An extreme evaluated sample remains evidence while its expected value is used in future references.","Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior."],"guidance":["Protection keeps the observed evidence; it does not delete observations or certify that a departure was an incident."],"path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","dependencies":["analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/robust.py","src/anomalyzer/baseline.py","src/anomalyzer/trend.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_trend.py","tests/test_cli.py"],"automatedChecks":[{"id":"robust-reference","description":"Limit an extreme weekly reference and use earlier references at lag one.","command":"python -m pytest -q tests/test_analysis.py::test_robust_weekly_reference_limits_one_high_week tests/test_analysis.py::test_robust_lag_one_uses_prior_seasonal_values"},{"id":"training-calibration","description":"Record training and calibration exclusions and retain a useful growth fit.","command":"python -m pytest -q tests/test_trend.py::test_robust_training_fit_and_calibration_exclusions"},{"id":"include-choice","description":"Compare robust and include modes through the CLI and show their differing triggers.","command":"python -m pytest -q tests/test_cli.py::test_outlier_handling_cli"}],"manualChecks":["Inspect excluded_from_model, reference_action, and outlier-handling diagnostics alongside the original observed value."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","criteria":[{"criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"covered","evidence":["src/anomalyzer/baseline.py:39","tests/test_analysis.py:25"],"notes":"Assertions check a independently computed weighted reference and invariance when an already-trimmed high value becomes more extreme."},{"criterion":"Training and calibration diagnostics identify excluded extreme observations.","status":"covered","evidence":["src/anomalyzer/trend.py:178","src/anomalyzer/recipes.py:287","tests/test_trend.py:65"],"notes":"Tests identify exact training/calibration exclusions and compare their influence on trend selection and residual scale."},{"criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"covered","evidence":["src/anomalyzer/recipes.py:378","src/anomalyzer/recipes.py:392","tests/test_analysis.py:263","tests/test_trend.py:11"],"notes":"The observed departure and trigger survive; only reference_values receives the prior expectation, preventing the next seasonal echo."},{"criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"covered","evidence":["src/anomalyzer/recipes.py:151","src/anomalyzer/recipes.py:475","tests/test_cli.py:95"],"notes":"CLI assertions distinguish robust triggers {65} from include-mode triggers {65, 72}."}],"ignore_findings":[],"done":true,"input_fingerprint":"183b8696fdc9786878b3e54c4a9e5f68180b47e157528bef953274a3487d49a7","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Protect later comparisons"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","label":"Choose the repeating cycle...","area":"analysis","scopes":["analysis","analysis/expectations"],"scope":"analysis/expectations","scopeLabel":"analysis/expectations","status":"implemented","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","acceptance":["An explicit season length is honored and reported as explicit.","A repeating positional series can infer a seasonal lag and report the selection window.","A positional series without a sufficiently stable candidate falls back to lag one and records why.","The reserved calibration and evaluation tail does not determine positional season selection."],"guidance":["Candidate discovery uses the input prefix before the reserved tail and confirms the candidate in an early window. It is not fully prefix-invariant streaming discovery.","Supply season_length explicitly when stable streaming season selection is required."],"path":".capabilities/analysis/expectations/select-season-length.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/seasonality.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"declared-season","description":"Honor a separate explicit season setting.","command":"python -m pytest -q tests/test_analysis.py::test_bare_values_with_separate_settings"},{"id":"inference-fallback","description":"Infer a repeating lag or record lag-one fallback when no season is supported.","command":"python -m pytest -q tests/test_analysis.py::test_positional_season_inference_and_fallback"},{"id":"reserved-tail","description":"Keep selection unchanged when only the reserved tail changes.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_season_inference_excludes_reserved_tail"},{"id":"cli-inference","description":"Expose the inferred lag and its source for bare JSON values in the CLI.","command":"python -m pytest -q tests/test_cli.py::test_bare_json_values_cli"}],"manualChecks":["Review selection_window and ranked candidates for a positional series before treating its inferred season as an ongoing monitoring assumption."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","criteria":[{"criterion":"An explicit season length is honored and reported as explicit.","status":"covered","evidence":["src/anomalyzer/recipes.py:55","tests/test_analysis.py:234"],"notes":"Explicit field presence bypasses inference, and the test checks the reported source."},{"criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"covered","evidence":["src/anomalyzer/seasonality.py:96","tests/test_analysis.py:247"],"notes":"The repeating seven-position series selects lag 7 and reports confirmation window [0, 27]."},{"criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"covered","evidence":["src/anomalyzer/seasonality.py:79","src/anomalyzer/seasonality.py:84","tests/test_analysis.py:247"],"notes":"Candidate-strength and early-confirmation failures return lag 1 with a reason; the quadratic example verifies fallback."},{"criterion":"The reserved calibration and evaluation tail does not determine positional season selection.","status":"covered","evidence":["src/anomalyzer/seasonality.py:69","tests/test_capability_behaviors.py:147"],"notes":"Discovery excludes calibration_size plus one reserved sample. Changing that tail preserves the complete inference diagnostics; arbitrary appended-prefix invariance is explicitly outside the guidance."}],"ignore_findings":[],"done":true,"input_fingerprint":"3317b1727dc3cabfcfc9b87808d9249d0d59f111b8fd5f8a1fad653b21343980","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","name":"Trend and seasonal analysis","label":"Trend and seasonal analysis / Refine what ordinary behavior means / Choose a repeating cycle"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","label":"Interpret rate changes wit...","area":"analysis","scopes":["analysis","analysis/rates"],"scope":"analysis/rates","scopeLabel":"analysis/rates","status":"planned","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","acceptance":["Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","Zero denominators remain unavailable, and low support is explicitly marked; intervals remain within the valid zero-to-one rate domain.","Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations."],"guidance":["Independence and stable exposure are modeling assumptions; clustered events, overdispersion, or changing populations require explicit limitations.","Proposed automated checks cover zero denominators, tiny counts, zero-or-all-success boundaries, and equal rates at different volumes.","Manual review should compare sparse and high-volume conversion changes and confirm the explanation discloses assumptions without inventing incident probability."],"path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Judge rates using their exposure"},"impact":0,"gaps":3,"r":47,"x":194,"y":524},{"id":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","label":"See changes before the day...","area":"analysis","scopes":["analysis","analysis/resolution"],"scope":"analysis/resolution","scopeLabel":"analysis/resolution","status":"implemented","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","acceptance":["Callers can select sum, mean, or last aggregation for completed and partial periods.","Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","An explicitly unfinished source sample is excluded from both analysis views.","A source cadence that does not evenly divide 24 hours is rejected."],"guidance":["Periods are fixed 24-hour windows, not daylight-saving-aware local calendar days.","Native and daily views share source observations and are correlated.","Legacy schema 1.0 requires complete input and does not support resets. Schema 1.1 labels sparse native/completed-day views and preserves reset boundaries; daily sums require declared additive volume and fully observed children."],"path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","dependencies":["data/quality/exclude-incomplete-periods","analysis/expectations/model-seasonal-level-and-growth"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/sparse_views.py","tests/test_sparse.py","src/anomalyzer/recipes.py","tests/test_analysis.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"selected-aggregation","description":"Check sum, mean, and last for completed days and trailing partial aggregates.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_multi_resolution_uses_selected_aggregation"},{"id":"two-views","description":"Return completed days, native evidence, and a linked trailing partial-day marker.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_current_period_and_keeps_two_views"},{"id":"unfinished-source","description":"Exclude a marked incomplete source sample from both views.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_marks_partial_source_sample_after_full_days"},{"id":"dividing-cadence","description":"Reject subperiod durations that do not evenly divide a day.","command":"python -m pytest -q tests/test_analysis.py::test_multi_resolution_rejects_non_dividing_frequency"}],"manualChecks":["Inspect the partial aggregate's observed_aggregate and intraday_evidence_refs together with the two method namespaces."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","criteria":[{"criterion":"Callers can select sum, mean, or last aggregation for completed and partial periods.","status":"covered","evidence":["src/anomalyzer/contracts.py:18","src/anomalyzer/recipes.py:539","tests/test_capability_behaviors.py:178"],"notes":"Parameterized assertions check independently known completed and partial values for all three aggregation choices."},{"criterion":"Regular timestamped sub-day samples produce distinct native and completed-period analysis views.","status":"covered","evidence":["src/anomalyzer/recipes.py:617","src/anomalyzer/recipes.py:667","src/anomalyzer/recipes.py:680","tests/test_analysis.py:170"],"notes":"Separate datasets, configurations, and namespaces preserve the two correlated views."},{"criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"covered","evidence":["src/anomalyzer/recipes.py:639","src/anomalyzer/recipes.py:690","tests/test_analysis.py:170"],"notes":"The marker contains finalized progress, observed aggregate, native evidence references, and an explicit exclusion flag."},{"criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"covered","evidence":["src/anomalyzer/validation.py:44","src/anomalyzer/recipes.py:564","tests/test_analysis.py:200"],"notes":"Both child datasets use prepared finalized observations; tests verify counts and retention of the excluded original value."},{"criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"covered","evidence":["src/anomalyzer/recipes.py:572","tests/test_analysis.py:221"],"notes":"A five-hour cadence raises the specific dividing-cadence error."}],"ignore_findings":[],"done":true,"input_fingerprint":"cdcdea1166a4e3d2523b594bec938142135e17a6b4279732e52819a486d1b4f2","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","name":"Incomplete-period monitoring","label":"Incomplete-period monitoring / Monitor while periods are still open / Compare finalized subperiods"},"impact":0,"gaps":0,"r":38,"x":194,"y":524},{"id":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","label":"Compare days and weeks in ...","area":"data","scopes":["data","data/calendar"],"scope":"data/calendar","scopeLabel":"data/calendar","status":"planned","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","acceptance":["Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","Periods report UTC boundaries and local labels; daylight-saving days can contain 23 or 25 hours without being forced into fixed 24-hour windows.","Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation."],"guidance":["Calendar grouping needs a distinct contract from fixed elapsed cadence; it must not silently reinterpret existing frequency or aggregation settings.","Proposed automated checks cover spring transitions, fall transitions, week boundaries, and missing native slots in a calendar period.","Manual review should inspect a daylight-saving week and confirm local labels, observed coverage, and completed-versus-open comparisons are understandable."],"path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Declare local calendar boundaries"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","label":"Analyze many entities with...","area":"data","scopes":["data","data/cohorts"],"scope":"data/cohorts","scopeLabel":"data/cohorts","status":"planned","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","acceptance":["Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","Training, calibration, relationships, and cases do not cross entity boundaries unless a separate explicit aggregate comparison is requested.","The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident."],"guidance":["Batch orchestration should preserve existing per-request contracts rather than silently lifting schema limits; multiple-testing limits stay visible.","Proposed automated checks cover partition isolation, batch-versus-single equivalence, one-entity failure, and shared resource-budget exhaustion.","Manual review should inspect a mixed batch of healthy, changed, and sparse entities and confirm the summary does not equate missing support with health."],"path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Keep each entity's evidence separate"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","label":"Know when missing updates ...","area":"data","scopes":["data","data/quality"],"scope":"data/quality","scopeLabel":"data/quality","status":"planned","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","acceptance":["Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","Results distinguish current, late, missing, and not-assessed freshness and expose the last observed update and the deadline used.","A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision."],"guidance":["Freshness depends on declared collection expectations; it does not prove source completeness and does not start a scheduler or send a notification.","Proposed automated checks cover exact deadlines, allowed lateness, absent source updates, and stale-source propagation to relationships.","Manual review should compare a stale feed with a true metric drop and confirm the report distinguishes collection uncertainty from numerical departure."],"path":".capabilities/data/quality/report-source-freshness.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","name":"Future - Calendar and data readiness","label":"Future - Calendar and data readiness / Monitor the periods my business uses / Check whether the sources are current"},"impact":0,"gaps":3,"r":47,"x":377.33333333333337,"y":524},{"id":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","label":"See how late or corrected ...","area":"data","scopes":["data","data/revisions"],"scope":"data/revisions","scopeLabel":"data/revisions","status":"planned","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","acceptance":["A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","Results preserve prior and new evidence versions and identify findings that were added, changed, or withdrawn with traceable source lineage.","Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results."],"guidance":["Schema-1.1 snapshot revision selection and as-of availability are implemented in data/quality/analyze-sparse-observations; durable prior/new evidence diffing and automatic reconciliation in this capability remain planned.","A corrected event-time result does not replace the record of what was known earlier; callers control retention and any notification about a revision.","Proposed automated checks cover late completion, corrected calibration data, downstream relationship changes, and conflicting source versions.","Manual review should inspect a drop withdrawn after ingestion completes and confirm both the original limitation and the revision explanation are retained."],"path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Reconcile revisions without losing history"},"impact":0,"gaps":3,"r":47,"x":353.33333333333337,"y":642},{"id":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","label":"Choose settings using held...","area":"evaluation","scopes":["evaluation","evaluation/comparisons"],"scope":"evaluation/comparisons","scopeLabel":"evaluation/comparisons","status":"planned","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","acceptance":["Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","The output preserves every candidate's settings and evaluation provenance; adopting a configuration requires an explicit caller decision."],"guidance":["Bounded comparison is not unrestricted parameter search; held-out results describe the supplied history and do not guarantee future effectiveness.","Proposed automated checks cover split isolation, comparable replay inputs, failed candidate retention, and unchanged active settings after comparison.","Manual review should compare an earlier but noisier configuration with a quieter one and confirm the tradeoff is visible without an unsupported winner."],"path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Review the tradeoff before adopting settings"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","label":"Measure detection timing a...","area":"evaluation","scopes":["evaluation","evaluation/outcomes"],"scope":"evaluation/outcomes","scopeLabel":"evaluation/outcomes","status":"planned","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","acceptance":["Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","Timing reports distinguish the incident's supplied start from the first available candidate, calibrated detection, and policy-eligible evidence.","Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown."],"guidance":["Detection timing does not measure response effectiveness or establish that an unmatched finding was false; incident labels are caller judgments.","Proposed automated checks cover boundary matching, ambiguous overlaps, incomplete label coverage, and separate candidate versus eligibility timing.","Manual review should inspect a matched and an unmatched episode and confirm the reported timing and accuracy limits agree with the supplied label coverage."],"path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Compare detection with reviewed outcomes"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","label":"Replay history as it would...","area":"evaluation","scopes":["evaluation","evaluation/replay"],"scope":"evaluation/replay","scopeLabel":"evaluation/replay","status":"planned","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","acceptance":["Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","Repeating a replay preserves numerical evidence and first-detection timing apart from generated IDs and runtime; stopped replays retain completed steps."],"guidance":["Event-time replay does not reconstruct arrival-time availability unless the caller supplies versioned arrival history.","Proposed automated checks cover future-tail isolation, short-history cutoffs, deterministic repeated replay, and interrupted replay retention.","Manual review should inspect one change from its first available observation through calibration and confirm the explanation uses only then-known evidence."],"path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","name":"Future - Historical evaluation","label":"Future - Historical evaluation / Evaluate a monitor before relying on it / Reconstruct the evidence available at the time"},"impact":0,"gaps":3,"r":47,"x":536.6666666666667,"y":642},{"id":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","label":"Update evidence when the n...","area":"execution","scopes":["execution","execution/incremental"],"scope":"execution/incremental","scopeLabel":"execution/incremental","status":"planned","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","acceptance":["Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","Supported explicitly configured recipes produce the same numerical evidence as full batch analysis of the combined history, including cross-chunk patterns.","Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations."],"guidance":["Start with explicit season lengths; input-prefix season discovery is not currently guaranteed to be invariant when the history grows.","Proposed automated checks cover chunk-versus-batch equivalence, cross-chunk CUSUM, idempotent retry, and failed-append state preservation.","Manual review should follow a sustained change across several append batches and confirm detection timing and provisional evidence remain understandable."],"path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Analyze the next finalized samples"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","label":"Keep useful evidence when ...","area":"execution","scopes":["execution","execution/resilience"],"scope":"execution/resilience","scopeLabel":"execution/resilience","status":"implemented","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","acceptance":["A method failure preserves evidence completed before the failure and reports the error.","Runtime budgets are checked cooperatively and produce an explicit stop reason.","CLI byte limits and execution failures produce distinct error exits.","A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial."],"guidance":["Default limits are 10,000 source observations per request, 10 MB CLI input, and 60 seconds cooperative runtime.","Numerical results are deterministic for fixed declared settings; generated run IDs and runtimes vary."],"path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","dependencies":["data/inputs/accept-analysis-requests"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/budget.py","src/anomalyzer/recipes.py","src/anomalyzer/orchestrator.py","src/anomalyzer/io.py","src/anomalyzer/cli.py","tests/test_analysis.py","tests/test_cli.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"method-and-budget","description":"Retain pre-failure evidence and report a runtime-budget stop.","command":"python -m pytest -q tests/test_analysis.py::test_failure_and_budget"},{"id":"cli-limits","description":"Enforce byte limits and distinguish execution failure in CLI output and exit status.","command":"python -m pytest -q tests/test_cli.py::test_size_limit_and_execution_exit"},{"id":"partial-siblings","description":"Preserve healthy analysis when a sibling dataset fails preparation.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_partial_analysis_preserves_healthy_siblings"}],"manualChecks":["Inspect run status, stop_reason, and each target's status before relying on a result; confirm limits are presented as cooperative rather than hard process timeouts."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","criteria":[{"criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"covered","evidence":["src/anomalyzer/recipes.py:400","src/anomalyzer/recipes.py:448","tests/test_analysis.py:509"],"notes":"The failure regression preserves evidence through index 44 and reports the injected error at index 45."},{"criterion":"Runtime budgets are checked cooperatively and produce an explicit stop reason.","status":"covered","evidence":["src/anomalyzer/budget.py:18","src/anomalyzer/recipes.py:108","src/anomalyzer/recipes.py:508","tests/test_analysis.py:509"],"notes":"Cooperative target/sample checks expose runtime_budget and budget_exceeded; this is not a hard process timeout."},{"criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"covered","evidence":["src/anomalyzer/io.py:9","src/anomalyzer/cli.py:429","tests/test_cli.py:108"],"notes":"The full suite executed the byte-limit and execution-exit regression successfully; invalid input exits 2 and execution failure exits 1."},{"criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"covered","evidence":["src/anomalyzer/orchestrator.py:125","src/anomalyzer/orchestrator.py:171","src/anomalyzer/orchestrator.py:185","tests/test_capability_behaviors.py:159","tests/test_capability_behaviors.py:30"],"notes":"Healthy dataset evidence survives target-specific failure; inspected incomplete-usage reports partial with no invented healthy relationship."}],"ignore_findings":[],"done":true,"input_fingerprint":"146a9c8396fd7a2d039f75077dc0b82a0778685d20c3782aa284d5f0b6ed94be","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Inspect incomplete execution"},"impact":0,"gaps":0,"r":38,"x":720,"y":642},{"id":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","label":"Resume local analysis afte...","area":"execution","scopes":["execution","execution/state"],"scope":"execution/state","scopeLabel":"execution/state","status":"planned","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","acceptance":["A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","Restoring a supported state and continuing yields the same numerical evidence and detection positions as uninterrupted processing of the same observations.","Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled."],"guidance":["Stored observations and context can contain sensitive data; document exactly what a snapshot contains and avoid automatic persistence.","Proposed automated checks cover restart equivalence, corrupt snapshots, version or settings mismatch, and output overwrite protection.","Manual review should inspect a snapshot and resumed explanation and confirm their provenance and retained-data scope are clear to an application owner."],"path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","name":"Future - Continuous analysis","label":"Future - Continuous analysis / Keep evidence current as observations arrive / Preserve state across restarts"},"impact":0,"gaps":3,"r":47,"x":720,"y":642},{"id":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","label":"Understand findings in a r...","area":"interfaces","scopes":["interfaces","interfaces/cli"],"scope":"interfaces/cli","scopeLabel":"interfaces/cli","status":"implemented","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","acceptance":["Point-anomaly episode counts and residual-pattern findings are labeled separately.","Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","Readiness status remains visible, including when it was not assessed."],"guidance":["Readable summaries should retain the evidence's claim limits and allow drilldown into JSON."],"path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/cli.py","tests/test_cli.py","examples/agentic/conversion.json","examples/agentic/unit-cost.json"],"automatedChecks":[{"id":"episodes-patterns","description":"Distinguish shift patterns from point episodes in readable output.","command":"python -m pytest -q tests/test_cli.py::test_readable_output_distinguishes_episodes_from_patterns"},{"id":"relationship-case","description":"Explain a conversion case with entity, units, comparison, and a suggested check.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_relationship_request_through_cli"},{"id":"explicit-rule","description":"Describe the unit-cost rule boundary and unassessed readiness.","command":"python -m pytest -q tests/test_cli.py::test_schema_11_readable_case_distinguishes_explicit_rule"}],"manualChecks":["Read the conversion and unit-cost example output and confirm the descriptions state the supported numerical comparison in plain language."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","criteria":[{"criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"covered","evidence":["src/anomalyzer/cli.py:347","src/anomalyzer/cli.py:405","tests/test_cli.py:143"],"notes":"Readable tests distinguish zero-episode residual shifts from a spike episode and retain the overlap warning."},{"criterion":"Relationship cases identify entity scope, observed and expected values, declared units, and investigation questions.","status":"covered","evidence":["src/anomalyzer/cli.py:291","src/anomalyzer/cli.py:308","src/anomalyzer/cli.py:331","tests/test_cli.py:117","examples/agentic/conversion.json"],"notes":"Conversion output identifies site=demo, observed/expected orders_per_visit, the relationship, and a collection-completeness investigation question."},{"criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"covered","evidence":["src/anomalyzer/cli.py:318","tests/test_cli.py:134","examples/agentic/unit-cost.json"],"notes":"Unit-cost output names review-above-13c and its [0, 0.13] boundary as a declared rule crossing."},{"criterion":"Readiness status remains visible, including when it was not assessed.","status":"covered","evidence":["src/anomalyzer/cli.py:352","src/anomalyzer/cli.py:399","tests/test_cli.py:134"],"notes":"Readable output includes not_assessed and its supporting reasons without claiming measured alert accuracy."}],"ignore_findings":[],"done":true,"input_fingerprint":"98108868b000972fd9b3c3f082846fbd3f3b08c399ef7ed424c23245d6935fef","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","name":"Basic CLI anomaly detection","label":"Basic CLI anomaly detection / Analyze a local series / Read the findings"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","label":"Share an investigation tha...","area":"interfaces","scopes":["interfaces","interfaces/export"],"scope":"interfaces/export","scopeLabel":"interfaces/export","status":"planned","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","acceptance":["Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","Source observations and context are included only by explicit selection; omitted data and resulting limits on independent reproduction are disclosed."],"guidance":["Fingerprints identify included provenance; they do not prove source truth. A portable report must make privacy scope and reproduction limits inspectable.","Proposed automated checks cover manifest consistency, partial-run export, omitted-source disclosure, and overwrite protection.","Manual review should open a bundle outside the original session and confirm a reviewer can trace one finding and see exactly what cannot be reproduced."],"path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Preserve a portable evidence record"},"impact":0,"gaps":3,"r":47,"x":927.3333333333334,"y":760},{"id":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","label":"Install a versioned analys...","area":"interfaces","scopes":["interfaces","interfaces/python"],"scope":"interfaces/python","scopeLabel":"interfaces/python","status":"implemented","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","acceptance":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"guidance":["Guard owns scheduling, persistence, business severity, incidents and notifications.","A published package does not establish integration into UsageTap Guard."],"path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","dependencies":["interfaces/python/analyze-through-public-api","interfaces/cli/analyze-files-and-streams","interfaces/mcp/connect-read-only-local-server"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["pyproject.toml","src/anomalyzer/_version.py","scripts/release.py","scripts/publish_release.py","scripts/verify_distribution.py","scripts/verify_lambda.sh","examples/library/guard_smoke.py",".github/workflows/ci.yml",".github/workflows/release.yml","docs/releases/README.md","tests/test_release.py"],"automatedChecks":[{"id":"release-safety","description":"Reject mismatched tags, altered assets and mutable publication.","command":"python -m pytest -q tests/test_release.py"},{"id":"installed-artifacts","description":"Install wheel and sdist outside the checkout and invoke public APIs.","command":"python scripts/verify_distribution.py --python 3.12"},{"id":"release-inventory","description":"Check archive contents, version metadata and SHA-256 inventory.","command":"python scripts/release.py verify"},{"id":"linux-offline-consumer","description":"Bundle and run with networking disabled in Linux Python 3.12.","command":"bash scripts/verify_lambda.sh"}],"manualChecks":["Follow the release guide and confirm Guard retains domain severity and incident ownership; inspect readiness and evidence limits before adopting."],"verificationGaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for interfaces/python/install-versioned-library.","criteria":[{"criterion":"Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","status":"partial","evidence":["src/anomalyzer/_version.py:1","scripts/release.py:1","scripts/release.py:20","scripts/release.py:39","scripts/release.py:48","scripts/release.py:54","scripts/publish_release.py:19","scripts/publish_release.py:27","scripts/publish_release.py:29","scripts/publish_release.py:39","scripts/verify_lambda.sh:14","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19","scripts/verify_lambda.sh:25","scripts/verify_lambda.sh:27",".github/workflows/ci.yml:14",".github/workflows/ci.yml:19",".github/workflows/ci.yml:27",".github/workflows/ci.yml:41",".github/workflows/ci.yml:46",".github/workflows/release.yml:28","docs/releases/README.md:12","docs/releases/README.md:13","docs/releases/README.md:14","docs/releases/README.md:17","docs/releases/README.md:18","tests/test_release.py:60","tests/test_release.py:71","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","status":"partial","evidence":["scripts/verify_distribution.py:1","scripts/verify_distribution.py:42","examples/library/guard_smoke.py:32","examples/library/guard_smoke.py:49","examples/library/guard_smoke.py:65","examples/library/guard_smoke.py:69","examples/library/guard_smoke.py:75",".github/workflows/ci.yml:57","docs/releases/README.md:13","docs/releases/README.md:208","docs/releases/README.md:211","docs/releases/README.md:213"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:34","scripts/release.py:81","scripts/release.py:82","scripts/release.py:85","scripts/release.py:92","scripts/publish_release.py:1","scripts/publish_release.py:25","scripts/publish_release.py:26","scripts/publish_release.py:27","scripts/publish_release.py:28","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:24","examples/library/guard_smoke.py:39",".github/workflows/ci.yml:42",".github/workflows/ci.yml:45",".github/workflows/ci.yml:46",".github/workflows/ci.yml:49",".github/workflows/ci.yml:51",".github/workflows/release.yml:31",".github/workflows/release.yml:33",".github/workflows/release.yml:36",".github/workflows/release.yml:37","docs/releases/README.md:3","docs/releases/README.md:12","docs/releases/README.md:14","docs/releases/README.md:15","docs/releases/README.md:19","tests/test_release.py:1","tests/test_release.py:21","tests/test_release.py:23","tests/test_release.py:52","tests/test_release.py:81"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","status":"partial","evidence":["pyproject.toml:26","pyproject.toml:33","scripts/release.py:81","scripts/release.py:82","scripts/publish_release.py:40","scripts/verify_distribution.py:41","scripts/verify_lambda.sh:6","scripts/verify_lambda.sh:17","scripts/verify_lambda.sh:19",".github/workflows/ci.yml:30",".github/workflows/ci.yml:31",".github/workflows/ci.yml:51",".github/workflows/ci.yml:52",".github/workflows/ci.yml:90","docs/releases/README.md:12","docs/releases/README.md:15","docs/releases/README.md:58","docs/releases/README.md:59","docs/releases/README.md:75"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Wheel and source distributions carry the complete Python package, license, public API, CLI and read-only MCP entry points with one authoritative version.","partial: Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","partial: Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","partial: A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled."],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"partial","storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","name":"Python integration","label":"Python integration / Embed analysis in my application / Install a released dependency"},"impact":0,"gaps":1,"r":41,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","label":"Ask an agent to explain un...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","acceptance":["The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit."],"guidance":["This contract covers the supplied instructions and demonstrated analysis workflow; it does not guarantee that every agent host follows the instructions.","Synthetic examples demonstrate mechanics, not real-world detection accuracy."],"path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams","interfaces/mcp/investigate-findings-and-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/agents/openai.yaml","src/anomalyzer/cli.py","src/anomalyzer/mcp.py","tests/test_cli.py","tests/test_agentic.py","tests/test_mcp.py"],"automatedChecks":[{"id":"structured-cli","description":"Exercise the documented structured CLI workflow and preserve incomplete-history evidence.","command":"python -m pytest -q tests/test_cli.py::test_stdin_file_equivalence tests/test_cli.py::test_bare_json_values_cli"},{"id":"declared-relationships","description":"Check explicit ratio semantics and unavailable source evidence.","command":"python -m pytest -q tests/test_agentic.py::test_ratio_lineage_and_early_relationship_departure"},{"id":"mcp-drilldown","description":"Follow compact analysis to a finding and its source evidence.","command":"python -m pytest -q tests/test_mcp.py::test_agent_can_drill_into_relationship_case"}],"manualChecks":["Follow the conversion example through the CLI and MCP drilldown; compare the Skill instructions with actual evidence and check that its proposed text separates calibrated findings, early candidates, readiness, and caller-owned actions."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","criteria":[{"criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:13",".agents/skills/anomaly-detection/SKILL.md:41","tests/test_cli.py:13"],"notes":"Executable resolution and JSON commands are explicit. Manual conversion and spike CLI runs use the installed numerical engine."},{"criterion":"The Skill instructs callers to preserve missing data, select one entity and unit, and explicitly declare compatible sources for related-metric analysis.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:24",".agents/skills/anomaly-detection/SKILL.md:32","tests/test_cli.py:51","tests/test_agentic.py:211"],"notes":"Instructions preserve scope, units, missingness, declared relationships, and exact alignment; existing runtime checks establish these prerequisites."},{"criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:101",".agents/skills/anomaly-detection/SKILL.md:134",".agents/skills/anomaly-detection/agents/openai.yaml:4"],"notes":"The supplied instructions lead with plain-language changes and then qualify maturity/readiness/overlap. This establishes instructional content, not measured agent compliance."},{"criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:54",".agents/skills/anomaly-detection/SKILL.md:174","src/anomalyzer/mcp.py:169","tests/test_mcp.py:54"],"notes":"The documented three-tool workflow matches executed conversion drilldown, with session-only run IDs and explicit read-only/caller-owned delivery limits."}],"ignore_findings":[],"done":true,"input_fingerprint":"751723214aa340b1227ec347c22f130749fcdbfd49327c21b4c27ef68d7f0341","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Get an evidence-grounded explanation"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","label":"See the evidence in a char...","area":"interfaces","scopes":["interfaces","interfaces/skill"],"scope":"interfaces/skill","scopeLabel":"interfaces/skill","status":"implemented","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","acceptance":["The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes."],"guidance":["Native chart rendering and interactive figure features depend on the agent host; the checked-in fallback produces a static SVG for a single dataset.","A chart explains structured analysis; it does not replace the detector result."],"path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","dependencies":["interfaces/cli/analyze-files-and-streams"],"suggestedDependencies":[],"dependents":[],"implementationReferences":[".agents/skills/anomaly-detection/SKILL.md",".agents/skills/anomaly-detection/scripts/quick_chart.py",".agents/skills/anomaly-detection/references/visualization.md","tests/test_skill_chart.py"],"automatedChecks":[{"id":"observed-evidence","description":"Render real observed and expected data and only actual point triggers.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_preserves_real_point_evidence"},{"id":"maturity-and-resets","description":"Keep provisional evidence non-triggering and label reviewed regimes.","command":"python -m pytest -q tests/test_skill_chart.py::test_provisional_scores_are_not_charted_as_triggers tests/test_skill_chart.py::test_quick_chart_marks_reviewed_resets"},{"id":"unsupported-scope","description":"Reject a multi-dataset result rather than mixing units.","command":"python -m pytest -q tests/test_skill_chart.py::test_quick_chart_rejects_multiple_datasets"},{"id":"output-protection","description":"Require an explicit overwrite for existing charts.","command":"python -m pytest -q tests/test_skill_chart.py::test_chart_cli_requires_explicit_overwrite"}],"manualChecks":["Inspect a spike fallback chart and the detailed figure instructions; verify that axes, initialization context, trigger markers, and interpretation limits are clear, and that the Skill requests a chart only when the user asks."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json.","Skill criteria covering agent instructions are supported by the written guidance and demonstrated tool workflow; cross-host agent compliance and real-world detection accuracy are not asserted."],"intent_summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","criteria":[{"criterion":"The Skill defaults to text and requires an explicit visualization request; quick charts and publication-quality figures have separate instructions.","status":"covered","evidence":[".agents/skills/anomaly-detection/SKILL.md:65",".agents/skills/anomaly-detection/SKILL.md:189",".agents/skills/anomaly-detection/SKILL.md:226"],"notes":"The Skill explicitly requires a visual request after analysis and separates quick versus detailed rendering instructions."},{"criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:81",".agents/skills/anomaly-detection/scripts/quick_chart.py:59","tests/test_skill_chart.py:28","tests/test_skill_chart.py:43","tests/test_skill_chart.py:54"],"notes":"Real-result tests establish observed/expected lines, trigger-only markers, provisional non-triggering, units, initialization shading, and reset boundaries; the spike chart was inspected visually."},{"criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"covered","evidence":[".agents/skills/anomaly-detection/scripts/quick_chart.py:85",".agents/skills/anomaly-detection/scripts/quick_chart.py:251","tests/test_skill_chart.py:63","tests/test_skill_chart.py:68"],"notes":"Multiple resolved datasets raise a specific error. The CLI test preserves an existing chart and replaces it only with explicit overwrite."},{"criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"covered","evidence":[".agents/skills/anomaly-detection/references/visualization.md:31",".agents/skills/anomaly-detection/references/visualization.md:37",".agents/skills/anomaly-detection/references/visualization.md:45",".agents/skills/anomaly-detection/references/visualization.md:86"],"notes":"Detailed instructions require index alignment, maturity distinctions, first detection timing, visible unavailable status, and qualified causal claims; native interactive rendering remains host-dependent."}],"ignore_findings":[],"done":true,"input_fingerprint":"2780b6c3e363f89a30a0709772c0a46a9a96f1567fabb27e63934a3ca80dc982","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","name":"Anomaly detection Skill","label":"Anomaly detection Skill / Ask an agent to investigate a change / Inspect a requested chart"},"impact":0,"gaps":0,"r":38,"x":927.3333333333334,"y":760},{"id":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","label":"Find which segments contri...","area":"investigation","scopes":["investigation","investigation/cohorts"],"scope":"investigation/cohorts","scopeLabel":"investigation/cohorts","status":"planned","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","acceptance":["Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","Segment ranking retains absolute signed contributions and baseline support; an arithmetic contributor is not automatically labeled anomalous or causal."],"guidance":["Ratios and overlapping customer groups are not additive totals; separate segment expectations must not be assumed to sum to a fitted total expectation.","Proposed automated checks cover exact reconciliation, offsetting segment changes, missing segments, and zero total departure.","Manual review should inspect a regional decline with an offsetting increase and confirm the explanation preserves signed contributions and evidence limits."],"path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Locate contributors to a total change"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","label":"Keep reviewed outcomes bes...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","acceptance":["A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","Review export distinguishes analyst judgment from detector output; annotation alone does not change calibration, eligibility, or regime boundaries."],"guidance":["Caller-supplied reviewer identity is attribution, not built-in authentication; storage, access control, response execution, and incident ownership stay with callers.","Proposed automated checks cover evidence-version linking, unresolved reviews, revision retention, and unchanged numerical evidence after annotation.","Manual review should inspect a disputed case with two review revisions and confirm judgments, uncertainties, and original findings remain distinguishable."],"path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Record the reviewed outcome"},"impact":0,"gaps":3,"r":47,"x":1110.6666666666667,"y":760},{"id":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","label":"Find earlier reviewed case...","area":"investigation","scopes":["investigation","investigation/history"],"scope":"investigation/history","scopeLabel":"investigation/history","status":"planned","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","acceptance":["Search is bounded to caller-selected records and explicit compatibility filters; it does not scan unrelated local files or require a model API.","Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match."],"guidance":["Similar historical evidence does not establish the same cause or authorize repeating a past response; caller-selected records may be incomplete or biased.","Proposed automated checks cover compatibility filtering, bounded corpus access, missing comparison context, and no-match behavior.","Manual review should compare superficially similar cases with different outcomes and confirm precedent is presented as investigation context."],"path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","name":"Future - Investigation memory","label":"Future - Investigation memory / Reuse what an investigation teaches us / Inspect comparable reviewed history"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","label":"Change action eligibility ...","area":"investigation","scopes":["investigation","investigation/policy"],"scope":"investigation/policy","scopeLabel":"investigation/policy","status":"implemented","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","acceptance":["Default policy leaves action and notification eligibility disabled.","Explicit-rule and calibrated-departure eligibility are controlled independently.","Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged."],"guidance":["Eligibility does not send notifications, execute actions, check readiness, or confer authority to act.","Monitoring schedules, delivery, reviewed incident state, and deduplication belong to the caller."],"path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","dependencies":["investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/agent.py","src/anomalyzer/cases.py","src/anomalyzer/contracts.py","tests/test_agentic.py","tests/test_capability_behaviors.py"],"automatedChecks":[{"id":"independent-notifications","description":"Replay notification and action choices without changing numerical evidence or the original result.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_separates_action_from_notification"},{"id":"calibrated-choice","description":"Enable calibrated departures only with the corresponding policy choice.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_replay_enables_calibrated_departures_only_when_requested"},{"id":"nontriggering-candidates","description":"Keep descriptive candidates ineligible even under a permissive policy.","command":"python -m pytest -q tests/test_capability_behaviors.py::test_policy_keeps_early_candidates_ineligible"}],"manualChecks":["Inspect the unit-cost example under default and allow_explicit_rules policies and compare eligibility with the unchanged numerical violation."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","criteria":[{"criterion":"Default policy leaves action and notification eligibility disabled.","status":"covered","evidence":["src/anomalyzer/contracts.py:329","tests/test_capability_behaviors.py:74"],"notes":"Default flags disable both eligibility types, including after replay resets policy."},{"criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"covered","evidence":["src/anomalyzer/agent.py:47","tests/test_capability_behaviors.py:123"],"notes":"Explicit-only permission does not enable a calibrated departure; the matching calibrated choice does."},{"criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"covered","evidence":["src/anomalyzer/agent.py:53","tests/test_capability_behaviors.py:101","tests/test_capability_behaviors.py:134"],"notes":"Notification-only replay enables a met rule while keeping action disabled; permissive settings still cannot enable candidates."},{"criterion":"Replaying policy preserves numerical evidence and stable case IDs, records changed case revisions, and leaves the original result unchanged.","status":"covered","evidence":["src/anomalyzer/agent.py:42","src/anomalyzer/cases.py:57","src/anomalyzer/cases.py:105","tests/test_capability_behaviors.py:101","tests/test_agentic.py:49"],"notes":"Assertions compare unchanged original serialization, methods, evidence references, stable IDs, and eligibility-dependent revision changes."}],"ignore_findings":[],"done":true,"input_fingerprint":"ee67bcc1ec1380b8d84a0556d838d648ae6cef3f4f9b9e34e1b71e90eb00aeae","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","name":"Evidence and review policy","label":"Evidence and review policy / Decide how much to rely on a finding / Review action eligibility"},"impact":0,"gaps":0,"r":38,"x":1086.6666666666667,"y":878},{"id":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","label":"Review a possible new base...","area":"investigation","scopes":["investigation","investigation/regimes"],"scope":"investigation/regimes","scopeLabel":"investigation/regimes","status":"planned","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","acceptance":["A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","Producing a proposal leaves the active regime and original findings unchanged.","A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect."],"guidance":["A proposed boundary is not a confirmed cause or a retrospectively available detection; adaptive reference updates remain a separate behavior.","Proposed automated checks cover unchanged active results, short new regimes, rejected proposals, and accepted explicit-reset equivalence.","Manual review should inspect a persistent outage-like drop and a known launch and confirm neither becomes ordinary behavior merely because it persists."],"path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","name":"Future - Business-aware expectations","label":"Future - Business-aware expectations / Make expectations fit the process / Review a proposed regime transition"},"impact":0,"gaps":3,"r":47,"x":1086.6666666666667,"y":878},{"id":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","label":"Distinguish changing perfo...","area":"relationships","scopes":["relationships","relationships/cohorts"],"scope":"relationships/cohorts","scopeLabel":"relationships/cohorts","status":"planned","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","acceptance":["Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","The report preserves aggregate and cohort rates, population weights, and the declared decomposition formula and reference-period convention.","Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes."],"guidance":["Decomposition is descriptive accounting; it does not estimate an intervention effect or establish that the comparison populations are causally equivalent.","Proposed automated checks cover stable rates with changing mix, stable mix with changing rates, offsetting cohort rates, and new or empty cohorts.","Manual review should inspect a Simpson's-paradox example and confirm the aggregate movement and cohort evidence can both be understood correctly."],"path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","dependencies":[],"suggestedDependencies":[],"dependents":[],"implementationReferences":[],"automatedChecks":[],"manualChecks":[],"verificationGaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}],"reviewFindings":[],"reviewStale":false,"health":"gap","healthLabel":"Needs action","coverage":"uncovered","storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","name":"Future - Cohort investigation","label":"Future - Cohort investigation / Investigate changes across populations / Separate performance changes from population mix"},"impact":0,"gaps":3,"r":47,"x":1270,"y":878},{"id":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","label":"Check whether a metric fol...","area":"relationships","scopes":["relationships","relationships/comparisons"],"scope":"relationships/comparisons","scopeLabel":"relationships/comparisons","status":"implemented","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","acceptance":["A supplied lag and coefficient use the predictor at the declared earlier sample position.","Derived values retain response and predictor source indexes and source cutoffs in lineage.","When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","Samples lacking the earlier predictor or awaiting coefficient training remain unavailable."],"guidance":["Lag is measured in samples and is never searched automatically.","Exactly one of coefficient or training_size is required."],"path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","dependencies":["relationships/quality/require-compatible-sources"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/relationships.py","src/anomalyzer/orchestrator.py","tests/test_agentic.py","tests/test_examples.py","examples/agentic/lagged-response.json"],"automatedChecks":[{"id":"supplied-response","description":"Use the declared lag and coefficient with correct predictor indexes.","command":"python -m pytest -q tests/test_agentic.py::test_lagged_response_uses_supplied_coefficient_and_prior_sample"},{"id":"frozen-coefficient","description":"Keep fitted coefficient and earlier lineage invariant when future samples arrive.","command":"python -m pytest -q tests/test_agentic.py::test_estimated_lag_coefficient_is_frozen_and_prefix_invariant"},{"id":"delayed-example","description":"Run the checked-in lagged-response demonstration against expected results.","command":"python -m pytest -q tests/test_examples.py::test_agentic_demo_checks_all_cases"},{"id":"pending-training","description":"Retain unavailable lineage until the full declared training prefix arrives, then freeze its coefficient.","command":"python -m pytest -q tests/test_agentic.py::test_lag_fit_waits_for_complete_training_prefix"}],"manualChecks":["Inspect the lagged-response example's leading unavailable values, coefficient parameters, and first usable derived sample."],"verificationGaps":[],"review":{"depth":"verified","source":"coding-agent","gaps":[],"evidence":["Primary Codex agent re-inspected declared implementation references and criterion evidence in the current session after the outcome-title and deliverable-lane refactor.",".venv/Scripts/python.exe -m pytest -q: 154 passed in 13.98s; .capabilities/reviews/deliverable-lanes/verification.json.","Executed and inspected CLI examples, time preparation, policy replay, open-day monitoring, and MCP conversion evidence drilldown; .capabilities/reviews/deliverable-lanes/manual-checks.json."],"intent_summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","criteria":[{"criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"covered","evidence":["src/anomalyzer/relationships.py:115","tests/test_agentic.py:254"],"notes":"Supplied lag/coefficient uses the declared earlier predictor position and preserves the expected response residual."},{"criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"covered","evidence":["src/anomalyzer/relationships.py:152","tests/test_agentic.py:254","examples/agentic/lagged-response.json"],"notes":"Available and pending lineage retains existing source indexes and timestamp cutoffs; missing predictors do not become zero."},{"criterion":"When requested, the coefficient is fitted once from a declared training prefix and remains frozen.","status":"covered","evidence":["src/anomalyzer/relationships.py:47","tests/test_agentic.py:294","tests/test_agentic.py:330"],"notes":"The coefficient is fitted once after the complete declared prefix; the independent 20/7 result and prefix-invariance assertions pass."},{"criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"covered","evidence":["src/anomalyzer/relationships.py:122","tests/test_agentic.py:330"],"notes":"Short prefixes retain null derived values and explicit unavailable reasons, while an unidentifiable completed fit has a visible error."}],"ignore_findings":[],"done":true,"input_fingerprint":"fc4516ef6ab903dfa4918268d97049aff04fbf9556620ccb8259ff830ba85785","mode":"quick"},"reviewFindings":["Saved review is stale; reassess changed capability or implementation evidence."],"reviewStale":true,"health":"review","healthLabel":"Needs review","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","name":"Related-metric analysis","label":"Related-metric analysis / Investigate relationships I declare / Check delayed responses"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878},{"id":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","label":"Inspect related metrics be...","area":"relationships","scopes":["relationships","relationships/resolution"],"scope":"relationships/resolution","scopeLabel":"relationships/resolution","status":"implemented","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed fixed UTC 24-hour additive aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","acceptance":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"guidance":["Daily ratios and differences are supported. Daily lagged, normalized-residual and joint-condition definitions remain explicitly inapplicable; their native evidence remains eligible. Local-calendar aggregation is separate work.","Initial support should use explicit relationships and compatible boundaries; cross-cadence joining or automatic resampling needs a separate contract.","Proposed automated checks cover ratio-of-totals arithmetic, missing sources, mismatched period boundaries, and open-period evidence separation.","Manual review should inspect a conversion decline within an unfinished day and verify the explanation separates native evidence from partial totals."],"path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","dependencies":["data/quality/analyze-sparse-observations","relationships/comparisons/compare-declared-metrics"],"suggestedDependencies":[],"dependents":[],"implementationReferences":["src/anomalyzer/sparse_views.py","src/anomalyzer/relationships.py","src/anomalyzer/agent.py","tests/test_sparse.py","examples/sparse/unfinished_daily.json","docs/sparse-data.md"],"automatedChecks":[{"id":"ratio-of-sums","description":"Compute daily ratios from compatible sums and preserve missing-child coverage.","command":"python -m pytest -q tests/test_sparse.py::test_complete_daily_ratio_divides_sums_and_missing_child_invalidates_only_its_view"},{"id":"view-policies","description":"Keep native/daily rules separate and preserve case identity on replay.","command":"python -m pytest -q tests/test_sparse.py::test_daily_rules_are_not_silently_reused_from_hourly_rules tests/test_sparse.py::test_multi_resolution_mcp_drilldown_selects_correct_view_and_replays_case_ids"},{"id":"view-limits","description":"Report unsupported daily lag semantics without discarding native evidence.","command":"python -m pytest -q tests/test_sparse.py::test_unsupported_daily_relationship_is_explicit_and_does_not_reinterpret_lag"}],"manualChecks":["Inspect progression chapter 22 and confirm incomplete progress is not presented as a completed total or independent corroboration."],"verificationGaps":[],"review":{"depth":"partial","source":"deterministic-assessment","gaps":["Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."],"evidence":[],"intent_summary":"Implementation evidence synchronized from deterministic assessment for relationships/resolution/compare-finalized-subperiods.","criteria":[{"criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:19","src/anomalyzer/sparse_views.py:22","src/anomalyzer/sparse_views.py:139","src/anomalyzer/sparse_views.py:143","src/anomalyzer/sparse_views.py:164","src/anomalyzer/agent.py:14","src/anomalyzer/agent.py:20","src/anomalyzer/agent.py:23","src/anomalyzer/agent.py:29","tests/test_sparse.py:18","tests/test_sparse.py:59","tests/test_sparse.py:87","tests/test_sparse.py:99","tests/test_sparse.py:117","docs/sparse-data.md:31","docs/sparse-data.md:127","docs/sparse-data.md:159","docs/sparse-data.md:199","docs/sparse-data.md:203"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:19","src/anomalyzer/sparse_views.py:75","src/anomalyzer/sparse_views.py:94","src/anomalyzer/sparse_views.py:122","src/anomalyzer/sparse_views.py:123","src/anomalyzer/relationships.py:7","src/anomalyzer/relationships.py:19","src/anomalyzer/relationships.py:33","src/anomalyzer/relationships.py:34","src/anomalyzer/relationships.py:36","src/anomalyzer/agent.py:34","tests/test_sparse.py:10","tests/test_sparse.py:40","tests/test_sparse.py:50","tests/test_sparse.py:60","tests/test_sparse.py:62","docs/sparse-data.md:17","docs/sparse-data.md:41","docs/sparse-data.md:68","docs/sparse-data.md:97","docs/sparse-data.md:99"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:70","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:170","src/anomalyzer/relationships.py:90","src/anomalyzer/relationships.py:173","tests/test_sparse.py:230","tests/test_sparse.py:259","tests/test_sparse.py:313","docs/sparse-data.md:41","docs/sparse-data.md:111","docs/sparse-data.md:194","docs/sparse-data.md:197","docs/sparse-data.md:198"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"criterion":"Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration.","status":"partial","evidence":["src/anomalyzer/sparse_views.py:1","src/anomalyzer/sparse_views.py:29","src/anomalyzer/sparse_views.py:82","src/anomalyzer/sparse_views.py:130","src/anomalyzer/sparse_views.py:159","src/anomalyzer/relationships.py:157","src/anomalyzer/relationships.py:192","src/anomalyzer/agent.py:20","src/anomalyzer/agent.py:47","tests/test_sparse.py:217","tests/test_sparse.py:247","tests/test_sparse.py:251","tests/test_sparse.py:264","tests/test_sparse.py:386","examples/sparse/unfinished_daily.json:208","docs/sparse-data.md:8","docs/sparse-data.md:11","docs/sparse-data.md:37","docs/sparse-data.md:41","docs/sparse-data.md:45"],"notes":"Related implementation text was found, but deterministic matching cannot prove the behavior. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"ignore_findings":[],"done":false},"reviewFindings":["review is not done","review depth is partial","Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration. Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit.","partial: Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","partial: Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","partial: Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","partial: Open-period progress remains separate from completed-period evidence and identifies coverage; correlated resolutions do not count as independent corroboration."],"reviewStale":false,"health":"gap","healthLabel":"Review findings","coverage":"partial","storyMap":{"release":"Related-metric analysis","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","name":"Related-metric analysis","label":"Related-metric analysis / Monitor the periods my business uses / Compare related metrics while the period is open"},"impact":0,"gaps":0,"r":38,"x":1270,"y":878}],"links":[{"source":"data/inputs/accept-analysis-requests","target":"analysis/evidence/calibrate-departures","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/cli/analyze-files-and-streams","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"analysis/expectations/model-seasonal-level-and-growth","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"interfaces/python/analyze-through-public-api","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"investigation/cases/group-related-evidence","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/identify-residual-patterns","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/mcp/investigate-findings-and-evidence","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"investigation/assessments/describe-early-changes","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"relationships/quality/require-compatible-sources","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/reset-declared-regimes","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"analysis/expectations/reset-declared-regimes","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"analysis/detection/identify-residual-patterns","target":"data/quality/analyze-sparse-observations","kind":"committed"},{"source":"data/quality/prepare-timestamped-series","target":"data/quality/exclude-incomplete-periods","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/mcp/connect-read-only-local-server","kind":"committed"},{"source":"investigation/assessments/describe-early-changes","target":"investigation/rules/check-declared-numerical-rules","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"relationships/comparisons/compare-declared-metrics","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/detection/report-point-anomaly-episodes","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/evidence/assess-reference-readiness","kind":"committed"},{"source":"analysis/evidence/calibrate-departures","target":"analysis/expectations/adapt-to-changing-growth","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/expectations/protect-reference-from-extremes","kind":"committed"},{"source":"data/quality/exclude-incomplete-periods","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"analysis/expectations/model-seasonal-level-and-growth","target":"analysis/resolution/monitor-within-unfinished-days","kind":"committed"},{"source":"data/inputs/accept-analysis-requests","target":"execution/resilience/retain-evidence-under-failure-and-limits","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"interfaces/cli/explain-findings-in-readable-output","kind":"committed"},{"source":"interfaces/python/analyze-through-public-api","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/mcp/connect-read-only-local-server","target":"interfaces/python/install-versioned-library","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/mcp/investigate-findings-and-evidence","target":"interfaces/skill/guide-anomaly-investigation","kind":"committed"},{"source":"interfaces/cli/analyze-files-and-streams","target":"interfaces/skill/visualize-analysis-on-request","kind":"committed"},{"source":"investigation/rules/check-declared-numerical-rules","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"investigation/cases/group-related-evidence","target":"investigation/policy/replay-action-eligibility","kind":"committed"},{"source":"relationships/quality/require-compatible-sources","target":"relationships/comparisons/evaluate-delayed-response","kind":"committed"},{"source":"data/quality/analyze-sparse-observations","target":"relationships/resolution/compare-finalized-subperiods","kind":"committed"},{"source":"relationships/comparisons/compare-declared-metrics","target":"relationships/resolution/compare-finalized-subperiods","kind":"committed"}],"scopes":["analysis","analysis/detection","analysis/evidence","analysis/expectations","analysis/rates","analysis/resolution","data","data/calendar","data/cohorts","data/inputs","data/quality","data/revisions","evaluation","evaluation/comparisons","evaluation/outcomes","evaluation/replay","execution","execution/incremental","execution/resilience","execution/state","interfaces","interfaces/cli","interfaces/export","interfaces/mcp","interfaces/python","interfaces/skill","investigation","investigation/assessments","investigation/cases","investigation/cohorts","investigation/history","investigation/policy","investigation/regimes","investigation/rules","relationships","relationships/cohorts","relationships/comparisons","relationships/quality","relationships/resolution"],"storyMaps":["Anomaly detection Skill","Basic CLI anomaly detection","Evidence and review policy","Future - Business-aware expectations","Future - Calendar and data readiness","Future - Cohort investigation","Future - Continuous analysis","Future - Historical evaluation","Future - Investigation memory","Incomplete-period monitoring","Python integration","Read-only MCP server","Related-metric analysis","Trend and seasonal analysis"],"hasUnassignedStoryMap":false};
 const width = 1440;
 const height = 980;
 const padding = 74;
diff --git a/.capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml b/.capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml
index 3fb6f31..bad9c30 100644
--- a/.capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml
+++ b/.capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml
@@ -59,7 +59,7 @@ agent:
         description: Inspect descriptive candidate evidence without inventing a case.
         command: python -m pytest -q
           tests/test_mcp.py::test_short_history_candidate_is_visible_without_a_case
-          tests/test_mcp.py::test_inapplicable_method_keeps_descriptive_candidate_visible
+          tests/test_mcp.py::test_sparse_insufficient_history_keeps_descriptive_candidate_visible
       - id: sample-pagination
         description: Page supporting samples for a residual pattern.
         command: python -m pytest -q
diff --git a/.capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml b/.capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml
index 338eaab..9215434 100644
--- a/.capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml
+++ b/.capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml
@@ -1,7 +1,7 @@
 title: Inspect related metrics before the business period closes
-status: planned
+status: implemented
 summary: Extend declared related-metric analysis to compatible finalized
-  subperiods and completed calendar or fixed-duration aggregates.
+  subperiods and completed fixed UTC 24-hour additive aggregates.
 intent: Detect conversion or unit-cost changes during a period while keeping
   partial progress separate from completed-period comparisons.
 acceptance:
@@ -16,6 +16,10 @@ acceptance:
     identifies coverage; correlated resolutions do not count as independent
     corroboration.
 guidance:
+  - Daily ratios and differences are supported. Daily lagged,
+    normalized-residual and joint-condition definitions remain explicitly
+    inapplicable; their native evidence remains eligible. Local-calendar
+    aggregation is separate work.
   - Initial support should use explicit relationships and compatible boundaries;
     cross-cadence joining or automatic resampling needs a separate contract.
   - Proposed automated checks cover ratio-of-totals arithmetic, missing sources,
@@ -26,5 +30,184 @@ planning:
   story_map:
     backbone: Monitor the periods my business uses
     step: Compare related metrics while the period is open
-    release: Future - Calendar and data readiness
+    release: Related-metric analysis
     order: 30
+#
+#
+agent:
+  # -----------------------------
+  # machine managed agent metadata
+  # review all capabilities and save evidence: capabilitykit review
+  # review this capability and save evidence: capabilitykit review relationships/resolution/compare-finalized-subperiods
+  # run deterministic review only: capabilitykit review relationships/resolution/compare-finalized-subperiods --deterministic-only
+  # ask an agent and save review evidence: capabilitykit review relationships/resolution/compare-finalized-subperiods --agent codex --arg exec --handoff stdin
+  # validate saved agent output without writing: capabilitykit review-result relationships/resolution/compare-finalized-subperiods --input review.json
+  # -----------------------------
+  depends_on:
+    - data/quality/analyze-sparse-observations
+    - relationships/comparisons/compare-declared-metrics
+  implementation:
+    references:
+      - src/anomalyzer/sparse_views.py
+      - src/anomalyzer/relationships.py
+      - src/anomalyzer/agent.py
+      - tests/test_sparse.py
+      - examples/sparse/unfinished_daily.json
+      - docs/sparse-data.md
+  verification:
+    automated:
+      - id: ratio-of-sums
+        description: Compute daily ratios from compatible sums and preserve
+          missing-child coverage.
+        command: python -m pytest -q
+          tests/test_sparse.py::test_complete_daily_ratio_divides_sums_and_missing_child_invalidates_only_its_view
+      - id: view-policies
+        description: Keep native/daily rules separate and preserve case identity on replay.
+        command: python -m pytest -q
+          tests/test_sparse.py::test_daily_rules_are_not_silently_reused_from_hourly_rules
+          tests/test_sparse.py::test_multi_resolution_mcp_drilldown_selects_correct_view_and_replays_case_ids
+      - id: view-limits
+        description: Report unsupported daily lag semantics without discarding native
+          evidence.
+        command: python -m pytest -q
+          tests/test_sparse.py::test_unsupported_daily_relationship_is_explicit_and_does_not_reinterpret_lag
+    manual:
+      - Inspect progression chapter 22 and confirm incomplete progress is not
+        presented as a completed total or independent corroboration.
+  review:
+    depth: partial
+    source: deterministic-assessment
+    gaps:
+      - Callers select related datasets with identical native boundaries, entity
+        scope, declared aggregation semantics, and compatible period
+        definitions. Review the cited evidence. If it is correct, save
+        agent.review evidence or make the criterion/reference wording more
+        explicit.
+      - Finalized subperiod relationships retain each source value and original
+        source index; unavailable or incomplete sources yield explicit
+        unavailability. Review the cited evidence. If it is correct, save
+        agent.review evidence or make the criterion/reference wording more
+        explicit.
+      - Completed aggregate ratios use aggregated numerators and denominators; a
+        ratio of totals is not silently replaced by an average of individual
+        ratios. Review the cited evidence. If it is correct, save agent.review
+        evidence or make the criterion/reference wording more explicit.
+      - Open-period progress remains separate from completed-period evidence and
+        identifies coverage; correlated resolutions do not count as independent
+        corroboration. Review the cited evidence. If it is correct, save
+        agent.review evidence or make the criterion/reference wording more
+        explicit.
+    intent_summary: Implementation evidence synchronized from deterministic
+      assessment for relationships/resolution/compare-finalized-subperiods.
+    criteria:
+      - criterion: Callers select related datasets with identical native boundaries,
+          entity scope, declared aggregation semantics, and compatible period
+          definitions.
+        status: partial
+        evidence:
+          - src/anomalyzer/sparse_views.py:19
+          - src/anomalyzer/sparse_views.py:22
+          - src/anomalyzer/sparse_views.py:139
+          - src/anomalyzer/sparse_views.py:143
+          - src/anomalyzer/sparse_views.py:164
+          - src/anomalyzer/agent.py:14
+          - src/anomalyzer/agent.py:20
+          - src/anomalyzer/agent.py:23
+          - src/anomalyzer/agent.py:29
+          - tests/test_sparse.py:18
+          - tests/test_sparse.py:59
+          - tests/test_sparse.py:87
+          - tests/test_sparse.py:99
+          - tests/test_sparse.py:117
+          - docs/sparse-data.md:31
+          - docs/sparse-data.md:127
+          - docs/sparse-data.md:159
+          - docs/sparse-data.md:199
+          - docs/sparse-data.md:203
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: Finalized subperiod relationships retain each source value and
+          original source index; unavailable or incomplete sources yield
+          explicit unavailability.
+        status: partial
+        evidence:
+          - src/anomalyzer/sparse_views.py:19
+          - src/anomalyzer/sparse_views.py:75
+          - src/anomalyzer/sparse_views.py:94
+          - src/anomalyzer/sparse_views.py:122
+          - src/anomalyzer/sparse_views.py:123
+          - src/anomalyzer/relationships.py:7
+          - src/anomalyzer/relationships.py:19
+          - src/anomalyzer/relationships.py:33
+          - src/anomalyzer/relationships.py:34
+          - src/anomalyzer/relationships.py:36
+          - src/anomalyzer/agent.py:34
+          - tests/test_sparse.py:10
+          - tests/test_sparse.py:40
+          - tests/test_sparse.py:50
+          - tests/test_sparse.py:60
+          - tests/test_sparse.py:62
+          - docs/sparse-data.md:17
+          - docs/sparse-data.md:41
+          - docs/sparse-data.md:68
+          - docs/sparse-data.md:97
+          - docs/sparse-data.md:99
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: Completed aggregate ratios use aggregated numerators and
+          denominators; a ratio of totals is not silently replaced by an average
+          of individual ratios.
+        status: partial
+        evidence:
+          - src/anomalyzer/sparse_views.py:29
+          - src/anomalyzer/sparse_views.py:70
+          - src/anomalyzer/sparse_views.py:82
+          - src/anomalyzer/sparse_views.py:170
+          - src/anomalyzer/relationships.py:90
+          - src/anomalyzer/relationships.py:173
+          - tests/test_sparse.py:230
+          - tests/test_sparse.py:259
+          - tests/test_sparse.py:313
+          - docs/sparse-data.md:41
+          - docs/sparse-data.md:111
+          - docs/sparse-data.md:194
+          - docs/sparse-data.md:197
+          - docs/sparse-data.md:198
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+      - criterion: Open-period progress remains separate from completed-period evidence
+          and identifies coverage; correlated resolutions do not count as
+          independent corroboration.
+        status: partial
+        evidence:
+          - src/anomalyzer/sparse_views.py:1
+          - src/anomalyzer/sparse_views.py:29
+          - src/anomalyzer/sparse_views.py:82
+          - src/anomalyzer/sparse_views.py:130
+          - src/anomalyzer/sparse_views.py:159
+          - src/anomalyzer/relationships.py:157
+          - src/anomalyzer/relationships.py:192
+          - src/anomalyzer/agent.py:20
+          - src/anomalyzer/agent.py:47
+          - tests/test_sparse.py:217
+          - tests/test_sparse.py:247
+          - tests/test_sparse.py:251
+          - tests/test_sparse.py:264
+          - tests/test_sparse.py:386
+          - examples/sparse/unfinished_daily.json:208
+          - docs/sparse-data.md:8
+          - docs/sparse-data.md:11
+          - docs/sparse-data.md:37
+          - docs/sparse-data.md:41
+          - docs/sparse-data.md:45
+        notes: Related implementation text was found, but deterministic matching cannot
+          prove the behavior. Review the cited evidence. If it is correct, save
+          agent.review evidence or make the criterion/reference wording more
+          explicit.
+    done: false
diff --git a/.capabilities/story-map-viewer.html b/.capabilities/story-map-viewer.html
index d53b42c..d54c2ec 100644
--- a/.capabilities/story-map-viewer.html
+++ b/.capabilities/story-map-viewer.html
@@ -492,7 +492,7 @@ <h1>Anomalyzer story map</h1>
       <dialog id="detail-dialog"></dialog>
     </div>
     <script>
-const report = {"project":"Anomalyzer","capabilities":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","health":"action","coverage":"partial","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","references":{"total":11,"readable":11,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Resolve or explicitly ignore verification gaps.","topFindings":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/publish_release.py","line":40,"excerpt":"\"Linux/Python 3.12 Lambda consumer verified offline. \""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":6,"excerpt":"bundle=\"$root/.tmp/lambda-consumer\""},{"reference":"scripts/verify_lambda.sh","line":17,"excerpt":"python -m pip download --only-binary=:all: --require-hashes \\"},{"reference":"scripts/verify_lambda.sh","line":19,"excerpt":"python -m pip install --no-index --only-binary=:all: --require-hashes \\"},{"reference":".github/workflows/ci.yml","line":30,"excerpt":"- run: uv run python examples/run_demo.py"},{"reference":".github/workflows/ci.yml","line":31,"excerpt":"- run: uv run python examples/run_agentic_demo.py"},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/ci.yml","line":52,"excerpt":"run: uv export --locked --no-dev --no-emit-project --output-file dist/runtime-dependencies.txt"},{"reference":".github/workflows/ci.yml","line":90,"excerpt":"name: Offline Lambda / Linux x86_64 / Python 3.12"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":58,"excerpt":"uv run python examples/run_demo.py"},{"reference":"docs/releases/README.md","line":59,"excerpt":"uv run python examples/run_agentic_demo.py"},{"reference":"docs/releases/README.md","line":75,"excerpt":"versions, and tests a bundled Linux x86_64/Python 3.12 consumer in AWS's Lambda"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","evidence":[{"reference":"scripts/verify_distribution.py","line":1,"excerpt":"\"\"\"Install wheel AND sdist in clean environments outside the source tree (requires uv).\"\"\""},{"reference":"scripts/verify_distribution.py","line":42,"excerpt":"run([\"uv\", \"pip\", \"install\", \"--python\", python, \"--no-deps\", artifact], outside, env=env)"},{"reference":"examples/library/guard_smoke.py","line":32,"excerpt":"analyze_relationships, get_case, replay_policy,"},{"reference":"examples/library/guard_smoke.py","line":49,"excerpt":"assert legacy.methods[0].diagnostics[\"detection_readiness\"]"},{"reference":"examples/library/guard_smoke.py","line":65,"excerpt":"\"relationships\": [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"cost\","},{"reference":"examples/library/guard_smoke.py","line":69,"excerpt":"result = analyze_relationships(request)"},{"reference":"examples/library/guard_smoke.py","line":76,"excerpt":"short = analyze_relationships({**request.model_dump(mode=\"json\"),"},{"reference":".github/workflows/ci.yml","line":57,"excerpt":"- name: Test built wheel and sdist outside checkout"},{"reference":"docs/releases/README.md","line":13,"excerpt":"| Source distribution (`.tar.gz`, or sdist) | Source needed to build the package. Useful for rebuilding and inspection; installation also needs the declared build backend. |"},{"reference":"docs/releases/README.md","line":208,"excerpt":"`analyze_series`, `analyze_relationships`, `get_case`, `replay_policy` (also"},{"reference":"docs/releases/README.md","line":211,"excerpt":"`anomalyzer.contracts`. See the self-contained"},{"reference":"docs/releases/README.md","line":213,"excerpt":"all APIs, recipes, point/pattern evidence, relationships, cases and policy"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":34,"excerpt":"return hashlib.sha256(path.read_bytes()).hexdigest()"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/release.py","line":85,"excerpt":"suffix = f\" --hash=sha256:{digest(wheel)}\\n\""},{"reference":"scripts/release.py","line":92,"excerpt":"assets = sorted(p for p in dist.iterdir() if p.is_file()"},{"reference":"scripts/publish_release.py","line":1,"excerpt":"\"\"\"Publish tested assets without replacing tags or existing assets (GitHub CLI auth).\"\"\""},{"reference":"scripts/publish_release.py","line":25,"excerpt":"remote_sha = json.loads(gh(\"api\", f\"repos/{REPOSITORY}/commits/{tag}\"))[\"sha\"]"},{"reference":"scripts/publish_release.py","line":26,"excerpt":"if remote_sha != revision:"},{"reference":"scripts/publish_release.py","line":27,"excerpt":"raise ValueError(\"Remote tag no longer points to the tested source revision\")"},{"reference":"scripts/publish_release.py","line":28,"excerpt":"endpoint = f\"repos/{REPOSITORY}/releases/tags/{tag}\""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":24,"excerpt":"sha256sum *.whl > /output/wheelhouse-SHA256SUMS"},{"reference":"examples/library/guard_smoke.py","line":39,"excerpt":"assert anomalyzer.__version__ == importlib.metadata.version(\"anomalyzer\")"},{"reference":".github/workflows/ci.yml","line":42,"excerpt":"- name: Check release tag before building"},{"reference":".github/workflows/ci.yml","line":45,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/ci.yml","line":46,"excerpt":"run: python scripts/release.py check-version --tag \"$RELEASE_TAG\""},{"reference":".github/workflows/ci.yml","line":49,"excerpt":"- name: Check package metadata"},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/release.yml","line":31,"excerpt":"name: release-dist-${{ github.sha }}"},{"reference":".github/workflows/release.yml","line":33,"excerpt":"- name: Stage tested assets for owner publication"},{"reference":".github/workflows/release.yml","line":36,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/release.yml","line":37,"excerpt":"run: python scripts/publish_release.py --tag \"$RELEASE_TAG\" --revision \"$GITHUB_SHA\""},{"reference":"docs/releases/README.md","line":3,"excerpt":"Anomalyzer is a normal Python dependency. Version **0.1.0** is the first release"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":14,"excerpt":"| Version tag (`v0.1.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":19,"excerpt":"it for package metadata and the public `__version__` imports it."},{"reference":"tests/test_release.py","line":1,"excerpt":"\"\"\"Release safety: reject wrong tags and refuse changes to published artifacts.\"\"\""},{"reference":"tests/test_release.py","line":21,"excerpt":"def test_release_refuses_mismatched_or_unsupported_tag(tag):"},{"reference":"tests/test_release.py","line":23,"excerpt":"release.check_tag(tag)"},{"reference":"tests/test_release.py","line":52,"excerpt":"def test_published_matching_release_is_noop(monkeypatch, tmp_path):"},{"reference":"tests/test_release.py","line":81,"excerpt":"\"version\": release.version(), \"source_revision\": \"a\" * 40, \"tag\": \"v0.1.0\","}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","order":20}},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","health":"review","coverage":"partial","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","order":50}},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","health":"review","coverage":"partial","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Readiness status remains visible, including when it was not assessed.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","order":60}},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","health":"review","coverage":"partial","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Analysis tool calls return structured content together with compact text.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","order":10}},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","order":20}},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","health":"review","coverage":"partial","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","order":10}},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","health":"review","coverage":"partial","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","order":10}},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","health":"review","coverage":"partial","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","order":20}},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","status":"planned","area":"analysis","path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","health":"planned","coverage":"uncovered","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","order":10}},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","status":"planned","area":"analysis","path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","health":"planned","coverage":"uncovered","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","order":20}},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","status":"planned","area":"data","path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","order":10}},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","status":"planned","area":"data","path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","health":"planned","coverage":"uncovered","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","order":10}},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","status":"planned","area":"data","path":".capabilities/data/quality/report-source-freshness.capability.yaml","health":"planned","coverage":"uncovered","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","order":20}},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","status":"planned","area":"data","path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","order":30}},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","status":"planned","area":"evaluation","path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","order":30}},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","status":"planned","area":"evaluation","path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","health":"planned","coverage":"uncovered","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","order":20}},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","status":"planned","area":"evaluation","path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","order":10}},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","status":"planned","area":"execution","path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","order":20}},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","status":"planned","area":"execution","path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","order":10}},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","status":"planned","area":"interfaces","path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","order":20}},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","status":"planned","area":"investigation","path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","order":20}},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","order":10}},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","health":"planned","coverage":"uncovered","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","order":30}},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","status":"planned","area":"investigation","path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","order":30}},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","status":"planned","area":"relationships","path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","order":30}},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","status":"planned","area":"relationships","path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed calendar or fixed-duration aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/resolution/compare-finalized-subperiods has no automated or manual verification checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/resolution/compare-finalized-subperiods has no automated checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-manual-review","message":"relationships/resolution/compare-finalized-subperiods has no manual review guidance","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","order":30}},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","health":"ok","coverage":"full","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","order":60}},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","health":"ok","coverage":"full","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","order":40}},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","health":"ok","coverage":"full","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","order":20}},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","health":"ok","coverage":"full","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","order":30}},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","health":"ok","coverage":"full","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","order":40}},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","health":"ok","coverage":"full","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","order":10}},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","health":"ok","coverage":"full","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","order":30}},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","health":"ok","coverage":"full","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","order":50}},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/select-season-length.capability.yaml","health":"ok","coverage":"full","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","order":20}},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","status":"implemented","area":"analysis","path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","health":"ok","coverage":"full","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":5,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","order":20}},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","status":"implemented","area":"data","path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","health":"ok","coverage":"full","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","order":10}},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","status":"implemented","area":"data","path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","health":"ok","coverage":"full","summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","order":10}},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","status":"implemented","area":"data","path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","health":"ok","coverage":"full","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","order":20}},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","status":"implemented","area":"execution","path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","health":"ok","coverage":"full","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","order":50}},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","status":"implemented","area":"investigation","path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","health":"ok","coverage":"full","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","order":10}},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","status":"implemented","area":"investigation","path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","health":"ok","coverage":"full","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","order":40}},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","status":"implemented","area":"investigation","path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","health":"ok","coverage":"full","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","order":60}},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","status":"implemented","area":"investigation","path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","health":"ok","coverage":"full","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","order":30}},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","health":"ok","coverage":"full","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","order":20}},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","health":"ok","coverage":"full","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","order":30}},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","status":"implemented","area":"relationships","path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","health":"ok","coverage":"full","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","order":10}}],"byStoryMap":{"unassigned":[],"releases":[{"release":"Anomaly detection Skill","capabilities":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","health":"review","coverage":"partial","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","order":10}},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","health":"review","coverage":"partial","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","order":20}}],"deliveryStrategy":{"release":"Anomaly detection Skill","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/skill/guide-anomaly-investigation"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/skill/guide-anomaly-investigation needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Ask an agent to investigate a change"],"missingBackbones":[],"stepCoverageGaps":["Ask an agent to investigate a change > Inspect a requested chart"],"rationale":"Starts with the earliest mapped step in Ask an agent to investigate a change."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/skill/visualize-analysis-on-request"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/skill/visualize-analysis-on-request needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Ask an agent to investigate a change"],"missingBackbones":[],"stepCoverageGaps":["Ask an agent to investigate a change > Get an evidence-grounded explanation"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/skill/visualize-analysis-on-request."}]},"presentation":{"outcome":"Anomaly detection Skill helps teams move from Get an evidence-grounded explanation to Inspect a requested chart across 1 backbone activity.","narrativePath":[{"backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","capabilityIds":["interfaces/skill/guide-anomaly-investigation"],"health":"review"},{"backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","capabilityIds":["interfaces/skill/visualize-analysis-on-request"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Ask an agent to investigate a change > Get an evidence-grounded explanation","message":"Ask an agent to investigate a change > Get an evidence-grounded explanation has needs-review coverage."},{"kind":"weak","label":"Ask an agent to investigate a change > Inspect a requested chart","message":"Ask an agent to investigate a change > Inspect a requested chart has needs-review coverage."}]}},{"release":"Basic CLI anomaly detection","capabilities":[{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","health":"ok","coverage":"full","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","order":40}},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","health":"ok","coverage":"full","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","order":30}},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","status":"implemented","area":"data","path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","health":"ok","coverage":"full","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","order":10}},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","status":"implemented","area":"data","path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","health":"ok","coverage":"full","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","order":20}},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","health":"review","coverage":"partial","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","order":50}},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","health":"review","coverage":"partial","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Readiness status remains visible, including when it was not assessed.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","order":60}}],"deliveryStrategy":{"release":"Basic CLI anomaly detection","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/inputs/accept-analysis-requests"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/inputs/accept-analysis-requests has normal delivery risk","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Analyze a local series"],"missingBackbones":[],"stepCoverageGaps":["Analyze a local series > Establish evidence support","Analyze a local series > Find spikes and drops","Analyze a local series > Prepare the comparison","Analyze a local series > Read the findings","Analyze a local series > Run from files or stdin"],"rationale":"Starts with the earliest mapped step in Analyze a local series."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["data/quality/prepare-timestamped-series","analysis/evidence/calibrate-departures","analysis/detection/report-point-anomaly-episodes"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/detection/report-point-anomaly-episodes has normal delivery risk; analysis/evidence/calibrate-departures has normal delivery risk; data/quality/prepare-timestamped-series has normal delivery risk","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Analyze a local series"],"missingBackbones":[],"stepCoverageGaps":["Analyze a local series > Read the findings","Analyze a local series > Run from files or stdin","Analyze a local series > Supply a numeric series"],"rationale":"Follows story-map order across Establish evidence support, Find spikes and drops, Prepare the comparison."},{"order":3,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/cli/analyze-files-and-streams","interfaces/cli/explain-findings-in-readable-output"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/cli/analyze-files-and-streams needs review; interfaces/cli/explain-findings-in-readable-output needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Analyze a local series"],"missingBackbones":[],"stepCoverageGaps":["Analyze a local series > Establish evidence support","Analyze a local series > Find spikes and drops","Analyze a local series > Prepare the comparison","Analyze a local series > Supply a numeric series"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/cli/analyze-files-and-streams, interfaces/cli/explain-findings-in-readable-output."}]},"presentation":{"outcome":"Basic CLI anomaly detection helps teams move from Supply a numeric series to Read the findings across 1 backbone activity.","narrativePath":[{"backbone":"Analyze a local series","step":"Supply a numeric series","capabilityIds":["data/inputs/accept-analysis-requests"],"health":"ok"},{"backbone":"Analyze a local series","step":"Prepare the comparison","capabilityIds":["data/quality/prepare-timestamped-series"],"health":"ok"},{"backbone":"Analyze a local series","step":"Establish evidence support","capabilityIds":["analysis/evidence/calibrate-departures"],"health":"ok"},{"backbone":"Analyze a local series","step":"Find spikes and drops","capabilityIds":["analysis/detection/report-point-anomaly-episodes"],"health":"ok"},{"backbone":"Analyze a local series","step":"Run from files or stdin","capabilityIds":["interfaces/cli/analyze-files-and-streams"],"health":"review"},{"backbone":"Analyze a local series","step":"Read the findings","capabilityIds":["interfaces/cli/explain-findings-in-readable-output"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Analyze a local series > Read the findings","message":"Analyze a local series > Read the findings has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Run from files or stdin","message":"Analyze a local series > Run from files or stdin has needs-review coverage."}]}},{"release":"Evidence and review policy","capabilities":[{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","health":"ok","coverage":"full","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","order":20}},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","status":"implemented","area":"execution","path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","health":"ok","coverage":"full","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","order":50}},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","status":"implemented","area":"investigation","path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","health":"ok","coverage":"full","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","order":10}},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","status":"implemented","area":"investigation","path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","health":"ok","coverage":"full","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","order":40}},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","status":"implemented","area":"investigation","path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","health":"ok","coverage":"full","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","order":60}},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","status":"implemented","area":"investigation","path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","health":"ok","coverage":"full","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","order":30}}],"deliveryStrategy":{"release":"Evidence and review policy","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["investigation/assessments/describe-early-changes"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"investigation/assessments/describe-early-changes has normal delivery risk","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Decide how much to rely on a finding"],"missingBackbones":[],"stepCoverageGaps":["Decide how much to rely on a finding > Apply my stated limits","Decide how much to rely on a finding > Assess baseline quality","Decide how much to rely on a finding > Inspect incomplete execution","Decide how much to rely on a finding > Investigate related evidence","Decide how much to rely on a finding > Review action eligibility"],"rationale":"Starts with the earliest mapped step in Decide how much to rely on a finding."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["analysis/evidence/assess-reference-readiness","investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence","execution/resilience/retain-evidence-under-failure-and-limits","investigation/policy/replay-action-eligibility"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/evidence/assess-reference-readiness has normal delivery risk; execution/resilience/retain-evidence-under-failure-and-limits has normal delivery risk; investigation/cases/group-related-evidence has normal delivery risk; investigation/policy/replay-action-eligibility has normal delivery risk; investigation/rules/check-declared-numerical-rules has normal delivery risk","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Decide how much to rely on a finding"],"missingBackbones":[],"stepCoverageGaps":["Decide how much to rely on a finding > Inspect early evidence"],"rationale":"Follows story-map order across Apply my stated limits, Assess baseline quality, Inspect incomplete execution, Investigate related evidence, Review action eligibility."}]},"presentation":{"outcome":"Evidence and review policy helps teams move from Inspect early evidence to Review action eligibility across 1 backbone activity.","narrativePath":[{"backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","capabilityIds":["investigation/assessments/describe-early-changes"],"health":"ok"},{"backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","capabilityIds":["analysis/evidence/assess-reference-readiness"],"health":"ok"},{"backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","capabilityIds":["investigation/rules/check-declared-numerical-rules"],"health":"ok"},{"backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","capabilityIds":["investigation/cases/group-related-evidence"],"health":"ok"},{"backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","capabilityIds":["execution/resilience/retain-evidence-under-failure-and-limits"],"health":"ok"},{"backbone":"Decide how much to rely on a finding","step":"Review action eligibility","capabilityIds":["investigation/policy/replay-action-eligibility"],"health":"ok"}],"coverageSignals":[]}},{"release":"Future - Business-aware expectations","capabilities":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","status":"planned","area":"analysis","path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","health":"planned","coverage":"uncovered","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","order":10}},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","status":"planned","area":"analysis","path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","health":"planned","coverage":"uncovered","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","order":20}},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","status":"planned","area":"investigation","path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","order":30}}],"deliveryStrategy":{"release":"Future - Business-aware expectations","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["analysis/expectations/model-multiple-seasonal-cycles"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"analysis/expectations/model-multiple-seasonal-cycles has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Make expectations fit the process"],"missingBackbones":[],"stepCoverageGaps":["Make expectations fit the process > Judge rates using their exposure","Make expectations fit the process > Review a proposed regime transition"],"rationale":"Starts with the earliest mapped step in Make expectations fit the process."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["analysis/rates/assess-bounded-count-rates","investigation/regimes/propose-reviewed-baseline-changes"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/rates/assess-bounded-count-rates has verification gaps; investigation/regimes/propose-reviewed-baseline-changes has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Make expectations fit the process"],"missingBackbones":[],"stepCoverageGaps":["Make expectations fit the process > Represent more than one repeating cycle"],"rationale":"Follows story-map order across Judge rates using their exposure, Review a proposed regime transition."}]},"presentation":{"outcome":"Future - Business-aware expectations helps teams move from Represent more than one repeating cycle to Review a proposed regime transition across 1 backbone activity.","narrativePath":[{"backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","capabilityIds":["analysis/expectations/model-multiple-seasonal-cycles"],"health":"planned"},{"backbone":"Make expectations fit the process","step":"Judge rates using their exposure","capabilityIds":["analysis/rates/assess-bounded-count-rates"],"health":"planned"},{"backbone":"Make expectations fit the process","step":"Review a proposed regime transition","capabilityIds":["investigation/regimes/propose-reviewed-baseline-changes"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Calendar and data readiness","capabilities":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","status":"planned","area":"data","path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","order":10}},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","status":"planned","area":"data","path":".capabilities/data/quality/report-source-freshness.capability.yaml","health":"planned","coverage":"uncovered","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","order":20}},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","status":"planned","area":"relationships","path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed calendar or fixed-duration aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/resolution/compare-finalized-subperiods has no automated or manual verification checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/resolution/compare-finalized-subperiods has no automated checks","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"},{"code":"missing-manual-review","message":"relationships/resolution/compare-finalized-subperiods has no manual review guidance","capabilityId":"relationships/resolution/compare-finalized-subperiods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\resolution\\compare-finalized-subperiods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/resolution/compare-finalized-subperiods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","order":30}}],"deliveryStrategy":{"release":"Future - Calendar and data readiness","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/calendar/aggregate-local-calendar-periods"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/calendar/aggregate-local-calendar-periods has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Monitor the periods my business uses"],"missingBackbones":[],"stepCoverageGaps":["Monitor the periods my business uses > Check whether the sources are current","Monitor the periods my business uses > Compare related metrics while the period is open"],"rationale":"Starts with the earliest mapped step in Monitor the periods my business uses."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["data/quality/report-source-freshness","relationships/resolution/compare-finalized-subperiods"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"data/quality/report-source-freshness has verification gaps; relationships/resolution/compare-finalized-subperiods has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Monitor the periods my business uses"],"missingBackbones":[],"stepCoverageGaps":["Monitor the periods my business uses > Declare local calendar boundaries"],"rationale":"Follows story-map order across Check whether the sources are current, Compare related metrics while the period is open."}]},"presentation":{"outcome":"Future - Calendar and data readiness helps teams move from Declare local calendar boundaries to Compare related metrics while the period is open across 1 backbone activity.","narrativePath":[{"backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","capabilityIds":["data/calendar/aggregate-local-calendar-periods"],"health":"planned"},{"backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","capabilityIds":["data/quality/report-source-freshness"],"health":"planned"},{"backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","capabilityIds":["relationships/resolution/compare-finalized-subperiods"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Cohort investigation","capabilities":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","status":"planned","area":"data","path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","health":"planned","coverage":"uncovered","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","order":10}},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","status":"planned","area":"investigation","path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","order":20}},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","status":"planned","area":"relationships","path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","order":30}}],"deliveryStrategy":{"release":"Future - Cohort investigation","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/cohorts/analyze-declared-entity-batches"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/cohorts/analyze-declared-entity-batches has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Investigate changes across populations"],"missingBackbones":[],"stepCoverageGaps":["Investigate changes across populations > Locate contributors to a total change","Investigate changes across populations > Separate performance changes from population mix"],"rationale":"Starts with the earliest mapped step in Investigate changes across populations."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["investigation/cohorts/explain-additive-segment-contributions","relationships/cohorts/separate-rate-and-population-mix"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"investigation/cohorts/explain-additive-segment-contributions has verification gaps; relationships/cohorts/separate-rate-and-population-mix has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Investigate changes across populations"],"missingBackbones":[],"stepCoverageGaps":["Investigate changes across populations > Keep each entity's evidence separate"],"rationale":"Follows story-map order across Locate contributors to a total change, Separate performance changes from population mix."}]},"presentation":{"outcome":"Future - Cohort investigation helps teams move from Keep each entity's evidence separate to Separate performance changes from population mix across 1 backbone activity.","narrativePath":[{"backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","capabilityIds":["data/cohorts/analyze-declared-entity-batches"],"health":"planned"},{"backbone":"Investigate changes across populations","step":"Locate contributors to a total change","capabilityIds":["investigation/cohorts/explain-additive-segment-contributions"],"health":"planned"},{"backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","capabilityIds":["relationships/cohorts/separate-rate-and-population-mix"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Continuous analysis","capabilities":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","status":"planned","area":"data","path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","order":30}},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","status":"planned","area":"execution","path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","order":20}},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","status":"planned","area":"execution","path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","order":10}}],"deliveryStrategy":{"release":"Future - Continuous analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["execution/state/save-and-restore-analysis-state"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"execution/state/save-and-restore-analysis-state has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Keep evidence current as observations arrive"],"missingBackbones":[],"stepCoverageGaps":["Keep evidence current as observations arrive > Analyze the next finalized samples","Keep evidence current as observations arrive > Reconcile revisions without losing history"],"rationale":"Starts with the earliest mapped step in Keep evidence current as observations arrive."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["execution/incremental/append-finalized-observations","data/revisions/reconcile-late-and-corrected-observations"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"data/revisions/reconcile-late-and-corrected-observations has verification gaps; execution/incremental/append-finalized-observations has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Keep evidence current as observations arrive"],"missingBackbones":[],"stepCoverageGaps":["Keep evidence current as observations arrive > Preserve state across restarts"],"rationale":"Follows story-map order across Analyze the next finalized samples, Reconcile revisions without losing history."}]},"presentation":{"outcome":"Future - Continuous analysis helps teams move from Preserve state across restarts to Reconcile revisions without losing history across 1 backbone activity.","narrativePath":[{"backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","capabilityIds":["execution/state/save-and-restore-analysis-state"],"health":"planned"},{"backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","capabilityIds":["execution/incremental/append-finalized-observations"],"health":"planned"},{"backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","capabilityIds":["data/revisions/reconcile-late-and-corrected-observations"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Historical evaluation","capabilities":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","status":"planned","area":"evaluation","path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","order":30}},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","status":"planned","area":"evaluation","path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","health":"planned","coverage":"uncovered","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","order":20}},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","status":"planned","area":"evaluation","path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","order":10}}],"deliveryStrategy":{"release":"Future - Historical evaluation","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["evaluation/replay/replay-historical-observations"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"evaluation/replay/replay-historical-observations has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Evaluate a monitor before relying on it"],"missingBackbones":[],"stepCoverageGaps":["Evaluate a monitor before relying on it > Compare detection with reviewed outcomes","Evaluate a monitor before relying on it > Review the tradeoff before adopting settings"],"rationale":"Starts with the earliest mapped step in Evaluate a monitor before relying on it."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["evaluation/outcomes/compare-findings-with-reviewed-incidents","evaluation/comparisons/compare-monitor-configurations"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"evaluation/comparisons/compare-monitor-configurations has verification gaps; evaluation/outcomes/compare-findings-with-reviewed-incidents has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Evaluate a monitor before relying on it"],"missingBackbones":[],"stepCoverageGaps":["Evaluate a monitor before relying on it > Reconstruct the evidence available at the time"],"rationale":"Follows story-map order across Compare detection with reviewed outcomes, Review the tradeoff before adopting settings."}]},"presentation":{"outcome":"Future - Historical evaluation helps teams move from Reconstruct the evidence available at the time to Review the tradeoff before adopting settings across 1 backbone activity.","narrativePath":[{"backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","capabilityIds":["evaluation/replay/replay-historical-observations"],"health":"planned"},{"backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","capabilityIds":["evaluation/outcomes/compare-findings-with-reviewed-incidents"],"health":"planned"},{"backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","capabilityIds":["evaluation/comparisons/compare-monitor-configurations"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Investigation memory","capabilities":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","status":"planned","area":"interfaces","path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","order":20}},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","order":10}},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","health":"planned","coverage":"uncovered","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","order":30}}],"deliveryStrategy":{"release":"Future - Investigation memory","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["investigation/history/record-reviewed-case-outcomes"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"investigation/history/record-reviewed-case-outcomes has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Reuse what an investigation teaches us"],"missingBackbones":[],"stepCoverageGaps":["Reuse what an investigation teaches us > Inspect comparable reviewed history","Reuse what an investigation teaches us > Preserve a portable evidence record"],"rationale":"Starts with the earliest mapped step in Reuse what an investigation teaches us."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["interfaces/export/package-investigation-evidence","investigation/history/retrieve-comparable-reviewed-cases"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"interfaces/export/package-investigation-evidence has verification gaps; investigation/history/retrieve-comparable-reviewed-cases has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Reuse what an investigation teaches us"],"missingBackbones":[],"stepCoverageGaps":["Reuse what an investigation teaches us > Record the reviewed outcome"],"rationale":"Follows story-map order across Inspect comparable reviewed history, Preserve a portable evidence record."}]},"presentation":{"outcome":"Future - Investigation memory helps teams move from Record the reviewed outcome to Inspect comparable reviewed history across 1 backbone activity.","narrativePath":[{"backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","capabilityIds":["investigation/history/record-reviewed-case-outcomes"],"health":"planned"},{"backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","capabilityIds":["interfaces/export/package-investigation-evidence"],"health":"planned"},{"backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","capabilityIds":["investigation/history/retrieve-comparable-reviewed-cases"],"health":"planned"}],"coverageSignals":[]}},{"release":"Incomplete-period monitoring","capabilities":[{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","status":"implemented","area":"analysis","path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","health":"ok","coverage":"full","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":5,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","order":20}},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","status":"implemented","area":"data","path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","health":"ok","coverage":"full","summary":"Explicitly incomplete trailing observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","order":10}}],"deliveryStrategy":{"release":"Incomplete-period monitoring","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/quality/exclude-incomplete-periods"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/quality/exclude-incomplete-periods has normal delivery risk","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Monitor while periods are still open"],"missingBackbones":[],"stepCoverageGaps":["Monitor while periods are still open > Compare finalized subperiods"],"rationale":"Starts with the earliest mapped step in Monitor while periods are still open."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["analysis/resolution/monitor-within-unfinished-days"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/resolution/monitor-within-unfinished-days has normal delivery risk","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Monitor while periods are still open"],"missingBackbones":[],"stepCoverageGaps":["Monitor while periods are still open > Keep unfinished data visible"],"rationale":"Follows story-map order across Compare finalized subperiods."}]},"presentation":{"outcome":"Incomplete-period monitoring helps teams move from Keep unfinished data visible to Compare finalized subperiods across 1 backbone activity.","narrativePath":[{"backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","capabilityIds":["data/quality/exclude-incomplete-periods"],"health":"ok"},{"backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","capabilityIds":["analysis/resolution/monitor-within-unfinished-days"],"health":"ok"}],"coverageSignals":[]}},{"release":"Python integration","capabilities":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","health":"review","coverage":"partial","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","order":10}},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","health":"action","coverage":"partial","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","references":{"total":11,"readable":11,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Resolve or explicitly ignore verification gaps.","topFindings":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/publish_release.py","line":40,"excerpt":"\"Linux/Python 3.12 Lambda consumer verified offline. \""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":6,"excerpt":"bundle=\"$root/.tmp/lambda-consumer\""},{"reference":"scripts/verify_lambda.sh","line":17,"excerpt":"python -m pip download --only-binary=:all: --require-hashes \\"},{"reference":"scripts/verify_lambda.sh","line":19,"excerpt":"python -m pip install --no-index --only-binary=:all: --require-hashes \\"},{"reference":".github/workflows/ci.yml","line":30,"excerpt":"- run: uv run python examples/run_demo.py"},{"reference":".github/workflows/ci.yml","line":31,"excerpt":"- run: uv run python examples/run_agentic_demo.py"},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/ci.yml","line":52,"excerpt":"run: uv export --locked --no-dev --no-emit-project --output-file dist/runtime-dependencies.txt"},{"reference":".github/workflows/ci.yml","line":90,"excerpt":"name: Offline Lambda / Linux x86_64 / Python 3.12"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":58,"excerpt":"uv run python examples/run_demo.py"},{"reference":"docs/releases/README.md","line":59,"excerpt":"uv run python examples/run_agentic_demo.py"},{"reference":"docs/releases/README.md","line":75,"excerpt":"versions, and tests a bundled Linux x86_64/Python 3.12 consumer in AWS's Lambda"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","evidence":[{"reference":"scripts/verify_distribution.py","line":1,"excerpt":"\"\"\"Install wheel AND sdist in clean environments outside the source tree (requires uv).\"\"\""},{"reference":"scripts/verify_distribution.py","line":42,"excerpt":"run([\"uv\", \"pip\", \"install\", \"--python\", python, \"--no-deps\", artifact], outside, env=env)"},{"reference":"examples/library/guard_smoke.py","line":32,"excerpt":"analyze_relationships, get_case, replay_policy,"},{"reference":"examples/library/guard_smoke.py","line":49,"excerpt":"assert legacy.methods[0].diagnostics[\"detection_readiness\"]"},{"reference":"examples/library/guard_smoke.py","line":65,"excerpt":"\"relationships\": [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"cost\","},{"reference":"examples/library/guard_smoke.py","line":69,"excerpt":"result = analyze_relationships(request)"},{"reference":"examples/library/guard_smoke.py","line":76,"excerpt":"short = analyze_relationships({**request.model_dump(mode=\"json\"),"},{"reference":".github/workflows/ci.yml","line":57,"excerpt":"- name: Test built wheel and sdist outside checkout"},{"reference":"docs/releases/README.md","line":13,"excerpt":"| Source distribution (`.tar.gz`, or sdist) | Source needed to build the package. Useful for rebuilding and inspection; installation also needs the declared build backend. |"},{"reference":"docs/releases/README.md","line":208,"excerpt":"`analyze_series`, `analyze_relationships`, `get_case`, `replay_policy` (also"},{"reference":"docs/releases/README.md","line":211,"excerpt":"`anomalyzer.contracts`. See the self-contained"},{"reference":"docs/releases/README.md","line":213,"excerpt":"all APIs, recipes, point/pattern evidence, relationships, cases and policy"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":34,"excerpt":"return hashlib.sha256(path.read_bytes()).hexdigest()"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/release.py","line":85,"excerpt":"suffix = f\" --hash=sha256:{digest(wheel)}\\n\""},{"reference":"scripts/release.py","line":92,"excerpt":"assets = sorted(p for p in dist.iterdir() if p.is_file()"},{"reference":"scripts/publish_release.py","line":1,"excerpt":"\"\"\"Publish tested assets without replacing tags or existing assets (GitHub CLI auth).\"\"\""},{"reference":"scripts/publish_release.py","line":25,"excerpt":"remote_sha = json.loads(gh(\"api\", f\"repos/{REPOSITORY}/commits/{tag}\"))[\"sha\"]"},{"reference":"scripts/publish_release.py","line":26,"excerpt":"if remote_sha != revision:"},{"reference":"scripts/publish_release.py","line":27,"excerpt":"raise ValueError(\"Remote tag no longer points to the tested source revision\")"},{"reference":"scripts/publish_release.py","line":28,"excerpt":"endpoint = f\"repos/{REPOSITORY}/releases/tags/{tag}\""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":24,"excerpt":"sha256sum *.whl > /output/wheelhouse-SHA256SUMS"},{"reference":"examples/library/guard_smoke.py","line":39,"excerpt":"assert anomalyzer.__version__ == importlib.metadata.version(\"anomalyzer\")"},{"reference":".github/workflows/ci.yml","line":42,"excerpt":"- name: Check release tag before building"},{"reference":".github/workflows/ci.yml","line":45,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/ci.yml","line":46,"excerpt":"run: python scripts/release.py check-version --tag \"$RELEASE_TAG\""},{"reference":".github/workflows/ci.yml","line":49,"excerpt":"- name: Check package metadata"},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/release.yml","line":31,"excerpt":"name: release-dist-${{ github.sha }}"},{"reference":".github/workflows/release.yml","line":33,"excerpt":"- name: Stage tested assets for owner publication"},{"reference":".github/workflows/release.yml","line":36,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/release.yml","line":37,"excerpt":"run: python scripts/publish_release.py --tag \"$RELEASE_TAG\" --revision \"$GITHUB_SHA\""},{"reference":"docs/releases/README.md","line":3,"excerpt":"Anomalyzer is a normal Python dependency. Version **0.1.0** is the first release"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":14,"excerpt":"| Version tag (`v0.1.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":19,"excerpt":"it for package metadata and the public `__version__` imports it."},{"reference":"tests/test_release.py","line":1,"excerpt":"\"\"\"Release safety: reject wrong tags and refuse changes to published artifacts.\"\"\""},{"reference":"tests/test_release.py","line":21,"excerpt":"def test_release_refuses_mismatched_or_unsupported_tag(tag):"},{"reference":"tests/test_release.py","line":23,"excerpt":"release.check_tag(tag)"},{"reference":"tests/test_release.py","line":52,"excerpt":"def test_published_matching_release_is_noop(monkeypatch, tmp_path):"},{"reference":"tests/test_release.py","line":81,"excerpt":"\"version\": release.version(), \"source_revision\": \"a\" * 40, \"tag\": \"v0.1.0\","}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","order":20}}],"deliveryStrategy":{"release":"Python integration","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/python/analyze-through-public-api"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/python/analyze-through-public-api needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Embed analysis in my application"],"missingBackbones":[],"stepCoverageGaps":["Embed analysis in my application > Install a released dependency"],"rationale":"Starts with the earliest mapped step in Embed analysis in my application."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/python/install-versioned-library"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/python/install-versioned-library has verification gaps; interfaces/python/install-versioned-library needs action","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Embed analysis in my application"],"missingBackbones":[],"stepCoverageGaps":["Embed analysis in my application > Call the public API"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/python/install-versioned-library."}]},"presentation":{"outcome":"Python integration helps teams move from Call the public API to Install a released dependency across 1 backbone activity.","narrativePath":[{"backbone":"Embed analysis in my application","step":"Call the public API","capabilityIds":["interfaces/python/analyze-through-public-api"],"health":"review"},{"backbone":"Embed analysis in my application","step":"Install a released dependency","capabilityIds":["interfaces/python/install-versioned-library"],"health":"action"}],"coverageSignals":[{"kind":"weak","label":"Embed analysis in my application > Call the public API","message":"Embed analysis in my application > Call the public API has needs-review coverage."},{"kind":"weak","label":"Embed analysis in my application > Install a released dependency","message":"Embed analysis in my application > Install a released dependency has needs-action coverage."}]}},{"release":"Read-only MCP server","capabilities":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","health":"review","coverage":"partial","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Analysis tool calls return structured content together with compact text.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","order":10}},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","order":20}}],"deliveryStrategy":{"release":"Read-only MCP server","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/mcp/connect-read-only-local-server"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/mcp/connect-read-only-local-server needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Give an agent access to local evidence"],"missingBackbones":[],"stepCoverageGaps":["Give an agent access to local evidence > Drill into findings and evidence"],"rationale":"Starts with the earliest mapped step in Give an agent access to local evidence."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/mcp/investigate-findings-and-evidence"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/mcp/investigate-findings-and-evidence needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Give an agent access to local evidence"],"missingBackbones":[],"stepCoverageGaps":["Give an agent access to local evidence > Connect a local server"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/mcp/investigate-findings-and-evidence."}]},"presentation":{"outcome":"Read-only MCP server helps teams move from Connect a local server to Drill into findings and evidence across 1 backbone activity.","narrativePath":[{"backbone":"Give an agent access to local evidence","step":"Connect a local server","capabilityIds":["interfaces/mcp/connect-read-only-local-server"],"health":"review"},{"backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","capabilityIds":["interfaces/mcp/investigate-findings-and-evidence"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Give an agent access to local evidence > Connect a local server","message":"Give an agent access to local evidence > Connect a local server has needs-review coverage."},{"kind":"weak","label":"Give an agent access to local evidence > Drill into findings and evidence","message":"Give an agent access to local evidence > Drill into findings and evidence has needs-review coverage."}]}},{"release":"Related-metric analysis","capabilities":[{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","health":"ok","coverage":"full","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","order":20}},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","health":"ok","coverage":"full","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","order":30}},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","status":"implemented","area":"relationships","path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","health":"ok","coverage":"full","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","order":10}}],"deliveryStrategy":{"release":"Related-metric analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["relationships/quality/require-compatible-sources"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"relationships/quality/require-compatible-sources has normal delivery risk","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Investigate relationships I declare"],"missingBackbones":[],"stepCoverageGaps":["Investigate relationships I declare > Check delayed responses","Investigate relationships I declare > Compare ratios and metric gaps"],"rationale":"Starts with the earliest mapped step in Investigate relationships I declare."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"relationships/comparisons/compare-declared-metrics has normal delivery risk; relationships/comparisons/evaluate-delayed-response has normal delivery risk","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Investigate relationships I declare"],"missingBackbones":[],"stepCoverageGaps":["Investigate relationships I declare > Align comparable sources"],"rationale":"Follows story-map order across Check delayed responses, Compare ratios and metric gaps."}]},"presentation":{"outcome":"Related-metric analysis helps teams move from Align comparable sources to Check delayed responses across 1 backbone activity.","narrativePath":[{"backbone":"Investigate relationships I declare","step":"Align comparable sources","capabilityIds":["relationships/quality/require-compatible-sources"],"health":"ok"},{"backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","capabilityIds":["relationships/comparisons/compare-declared-metrics"],"health":"ok"},{"backbone":"Investigate relationships I declare","step":"Check delayed responses","capabilityIds":["relationships/comparisons/evaluate-delayed-response"],"health":"ok"}],"coverageSignals":[]}},{"release":"Trend and seasonal analysis","capabilities":[{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","health":"ok","coverage":"full","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","order":60}},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","health":"ok","coverage":"full","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","order":40}},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","health":"ok","coverage":"full","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","order":10}},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","health":"ok","coverage":"full","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","order":30}},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","health":"ok","coverage":"full","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","order":50}},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/select-season-length.capability.yaml","health":"ok","coverage":"full","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":4,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"No immediate action.","topFindings":[],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","order":20}}],"deliveryStrategy":{"release":"Trend and seasonal analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["analysis/expectations/model-seasonal-level-and-growth"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"analysis/expectations/model-seasonal-level-and-growth has normal delivery risk","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Refine what ordinary behavior means"],"missingBackbones":[],"stepCoverageGaps":["Refine what ordinary behavior means > Choose a repeating cycle","Refine what ordinary behavior means > Find sustained residual changes","Refine what ordinary behavior means > Follow evolving growth","Refine what ordinary behavior means > Protect later comparisons","Refine what ordinary behavior means > Restart after a reviewed change"],"rationale":"Starts with the earliest mapped step in Refine what ordinary behavior means."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["analysis/expectations/select-season-length","analysis/expectations/protect-reference-from-extremes","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","analysis/detection/identify-residual-patterns"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/detection/identify-residual-patterns has normal delivery risk; analysis/expectations/adapt-to-changing-growth has normal delivery risk; analysis/expectations/protect-reference-from-extremes has normal delivery risk; analysis/expectations/reset-declared-regimes has normal delivery risk; analysis/expectations/select-season-length has normal delivery risk","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Refine what ordinary behavior means"],"missingBackbones":[],"stepCoverageGaps":["Refine what ordinary behavior means > Account for growth and seasonality"],"rationale":"Follows story-map order across Choose a repeating cycle, Find sustained residual changes, Follow evolving growth, Protect later comparisons, Restart after a reviewed change."}]},"presentation":{"outcome":"Trend and seasonal analysis helps teams move from Account for growth and seasonality to Find sustained residual changes across 1 backbone activity.","narrativePath":[{"backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","capabilityIds":["analysis/expectations/model-seasonal-level-and-growth"],"health":"ok"},{"backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","capabilityIds":["analysis/expectations/select-season-length"],"health":"ok"},{"backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","capabilityIds":["analysis/expectations/protect-reference-from-extremes"],"health":"ok"},{"backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","capabilityIds":["analysis/expectations/adapt-to-changing-growth"],"health":"ok"},{"backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","capabilityIds":["analysis/expectations/reset-declared-regimes"],"health":"ok"},{"backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","capabilityIds":["analysis/detection/identify-residual-patterns"],"health":"ok"}],"coverageSignals":[]}}]},"summary":{"total":47,"ok":21,"review":7,"action":1,"planned":18}};
+const report = {"project":"Anomalyzer","capabilities":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","health":"action","coverage":"partial","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","references":{"total":11,"readable":11,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Resolve or explicitly ignore verification gaps.","topFindings":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/publish_release.py","line":43,"excerpt":"\"Linux/Python 3.12 Lambda consumer verified offline. \""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":6,"excerpt":"bundle=\"$root/.tmp/lambda-consumer\""},{"reference":"scripts/verify_lambda.sh","line":18,"excerpt":"python -m pip download --only-binary=:all: --require-hashes \\"},{"reference":"scripts/verify_lambda.sh","line":20,"excerpt":"python -m pip install --no-index --only-binary=:all: --require-hashes \\"},{"reference":".github/workflows/ci.yml","line":29,"excerpt":"- run: uv run python examples/generate.py"},{"reference":".github/workflows/ci.yml","line":32,"excerpt":"- run: uv run python examples/run_demo.py"},{"reference":".github/workflows/ci.yml","line":33,"excerpt":"- run: uv run python examples/run_agentic_demo.py"},{"reference":".github/workflows/ci.yml","line":53,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/ci.yml","line":54,"excerpt":"run: uv export --locked --no-dev --no-emit-project --output-file dist/runtime-dependencies.txt"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":58,"excerpt":"uv run python examples/run_demo.py"},{"reference":"docs/releases/README.md","line":59,"excerpt":"uv run python examples/run_agentic_demo.py"},{"reference":"docs/releases/README.md","line":75,"excerpt":"versions, and tests a bundled Linux x86_64/Python 3.12 consumer in AWS's Lambda"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","evidence":[{"reference":"scripts/verify_distribution.py","line":1,"excerpt":"\"\"\"Install wheel AND sdist in clean environments outside the source tree (requires uv).\"\"\""},{"reference":"scripts/verify_distribution.py","line":42,"excerpt":"run([\"uv\", \"pip\", \"install\", \"--python\", python, \"--no-deps\", artifact], outside, env=env)"},{"reference":"examples/library/guard_smoke.py","line":32,"excerpt":"analyze_relationships, get_case, replay_policy,"},{"reference":"examples/library/guard_smoke.py","line":49,"excerpt":"assert legacy.methods[0].diagnostics[\"detection_readiness\"]"},{"reference":"examples/library/guard_smoke.py","line":65,"excerpt":"\"relationships\": [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"cost\","},{"reference":"examples/library/guard_smoke.py","line":69,"excerpt":"result = analyze_relationships(request)"},{"reference":"examples/library/guard_smoke.py","line":76,"excerpt":"short = analyze_relationships({**request.model_dump(mode=\"json\"),"},{"reference":".github/workflows/ci.yml","line":59,"excerpt":"- name: Test built wheel and sdist outside checkout"},{"reference":"docs/releases/README.md","line":13,"excerpt":"| Source distribution (`.tar.gz`, or sdist) | Source needed to build the package. Useful for rebuilding and inspection; installation also needs the declared build backend. |"},{"reference":"docs/releases/README.md","line":208,"excerpt":"`analyze_series`, `analyze_relationships`, `get_case`, `replay_policy` (also"},{"reference":"docs/releases/README.md","line":211,"excerpt":"`anomalyzer.contracts`. See the self-contained"},{"reference":"docs/releases/README.md","line":213,"excerpt":"all APIs, recipes, point/pattern evidence, relationships, cases and policy"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":34,"excerpt":"return hashlib.sha256(path.read_bytes()).hexdigest()"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/release.py","line":85,"excerpt":"suffix = f\" --hash=sha256:{digest(wheel)}\\n\""},{"reference":"scripts/release.py","line":92,"excerpt":"assets = sorted(p for p in dist.iterdir() if p.is_file()"},{"reference":"scripts/publish_release.py","line":1,"excerpt":"\"\"\"Publish tested assets without replacing tags or existing assets (GitHub CLI auth).\"\"\""},{"reference":"scripts/publish_release.py","line":25,"excerpt":"remote_sha = json.loads(gh(\"api\", f\"repos/{REPOSITORY}/commits/{tag}\"))[\"sha\"]"},{"reference":"scripts/publish_release.py","line":26,"excerpt":"if remote_sha != revision:"},{"reference":"scripts/publish_release.py","line":27,"excerpt":"raise ValueError(\"Remote tag no longer points to the tested source revision\")"},{"reference":"scripts/publish_release.py","line":28,"excerpt":"endpoint = f\"repos/{REPOSITORY}/releases/tags/{tag}\""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":25,"excerpt":"sha256sum *.whl > /output/wheelhouse-SHA256SUMS"},{"reference":"examples/library/guard_smoke.py","line":39,"excerpt":"assert anomalyzer.__version__ == importlib.metadata.version(\"anomalyzer\")"},{"reference":".github/workflows/ci.yml","line":44,"excerpt":"- name: Check release tag before building"},{"reference":".github/workflows/ci.yml","line":47,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/ci.yml","line":48,"excerpt":"run: python scripts/release.py check-version --tag \"$RELEASE_TAG\""},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Check package metadata"},{"reference":".github/workflows/ci.yml","line":53,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/release.yml","line":31,"excerpt":"name: release-dist-${{ github.sha }}"},{"reference":".github/workflows/release.yml","line":33,"excerpt":"- name: Stage tested assets for owner publication"},{"reference":".github/workflows/release.yml","line":36,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/release.yml","line":37,"excerpt":"run: python scripts/publish_release.py --tag \"$RELEASE_TAG\" --revision \"$GITHUB_SHA\""},{"reference":"docs/releases/README.md","line":3,"excerpt":"Anomalyzer is a normal Python dependency. Version **0.2.0** is the first release"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":14,"excerpt":"| Version tag (`v0.2.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":19,"excerpt":"it for package metadata and the public `__version__` imports it."},{"reference":"tests/test_release.py","line":1,"excerpt":"\"\"\"Release safety: reject wrong tags and refuse changes to published artifacts.\"\"\""},{"reference":"tests/test_release.py","line":21,"excerpt":"def test_release_refuses_mismatched_or_unsupported_tag(tag):"},{"reference":"tests/test_release.py","line":23,"excerpt":"release.check_tag(tag)"},{"reference":"tests/test_release.py","line":52,"excerpt":"def test_published_matching_release_is_noop(monkeypatch, tmp_path):"},{"reference":"tests/test_release.py","line":81,"excerpt":"\"version\": release.version(), \"source_revision\": \"a\" * 40, \"tag\": f\"v{release.version()}\","}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","order":20}},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","health":"review","coverage":"partial","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"Moving-range findings identify variation changes separately from location changes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","order":60}},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","health":"review","coverage":"partial","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","order":40}},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","health":"review","coverage":"partial","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","order":20}},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","health":"review","coverage":"partial","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","order":30}},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","health":"review","coverage":"partial","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","order":40}},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","health":"review","coverage":"partial","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","order":10}},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","health":"review","coverage":"partial","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","order":30}},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","health":"review","coverage":"partial","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","order":50}},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/select-season-length.capability.yaml","health":"review","coverage":"partial","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"An explicit season length is honored and reported as explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","order":20}},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","status":"implemented","area":"analysis","path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","health":"review","coverage":"partial","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","order":20}},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","status":"implemented","area":"data","path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","health":"review","coverage":"partial","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","order":10}},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","status":"implemented","area":"data","path":".capabilities/data/quality/analyze-sparse-observations.capability.yaml","health":"review","coverage":"partial","summary":"Schema-1.1 callers declare coverage and missing-data semantics per source, keeping unknown collection distinct from confirmed empty additive volume.","intent":"Retain useful rule and supported statistical evidence for sparse histories while preserving time, provenance, incomplete periods and evidence limits.","references":{"total":10,"readable":10,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Run semantic review, save review evidence, or ignore accepted findings.","topFindings":[{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":13,"excerpt":"\"seasonal-residual-v1\", \"adaptive-seasonal-v1\", \"multi-resolution-v1\","},{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":297,"excerpt":"if (self.data_policy or self.coverage or self.available_at or self.revisions) and self.timestamps is None:"},{"reference":"src/anomalyzer/contracts.py","line":298,"excerpt":"raise ValueError(\"source policies/coverage/revisions require original timestamps\")"},{"reference":"src/anomalyzer/contracts.py","line":437,"excerpt":"total_points = sum(len(dataset.values) + len(dataset.revisions) for dataset in self.datasets)"},{"reference":"src/anomalyzer/sparse.py","line":4,"excerpt":"Event coordinates retain actual timestamps and represent comparable events, not time."},{"reference":"src/anomalyzer/sparse.py","line":34,"excerpt":"if ds.data_policy or ds.coverage is not None or ds.available_at is not None or ds.revisions or any(v is None for v in ds.values) or \"incomplete\" in (ds.period_statuses or []):"},{"reference":"src/anomalyzer/sparse.py","line":64,"excerpt":"raise ValueError(\"duplicate timestamps require declared duplicate_policy; event_index needs unique event timestamps\")"},{"reference":"src/anomalyzer/sparse.py","line":66,"excerpt":"if coverage != \"complete\" or previous[\"coverage\"] != \"complete\" or ds.available_at or ds.revisions:"},{"reference":"src/anomalyzer/sparse.py","line":83,"excerpt":"observed_at = timestamp(revision.timestamp, ds.timezone)"},{"reference":"src/anomalyzer/sparse_views.py","line":1,"excerpt":"\"\"\"Schema-1.1 native and complete-day evidence with independent view eligibility.\"\"\""},{"reference":"src/anomalyzer/sparse_views.py","line":28,"excerpt":"case.revision_id = f\"{resolution}:{case.revision_id}\""},{"reference":"src/anomalyzer/sparse_views.py","line":29,"excerpt":"case.non_claims.append(\"Native and completed-day views reuse the same observations; they are not independent confirmation.\")"},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/orchestrator.py","line":106,"excerpt":"if request.config.recipe == \"multi-resolution-v1\":"},{"reference":"src/anomalyzer/orchestrator.py","line":107,"excerpt":"from .sparse_views import analyze_multi_request"},{"reference":"src/anomalyzer/orchestrator.py","line":133,"excerpt":"if \"data_policy\" in source.quality and request.config.recipe != \"multi-resolution-v1\""},{"reference":"src/anomalyzer/orchestrator.py","line":166,"excerpt":"if any(definitions[name].timestamps is None for name in _source_ids(relationship)):"},{"reference":"src/anomalyzer/orchestrator.py","line":180,"excerpt":"reset_positions=active_resets(request.config.reset_points, derived.timestamps or [], derived.timezone, request.context.as_of),"},{"reference":"src/anomalyzer/relationships.py","line":160,"excerpt":"source_timestamps = {source.dataset.id: source.times[source_index]"},{"reference":"src/anomalyzer/relationships.py","line":162,"excerpt":"cutoff = max(source_timestamps.values())"},{"reference":"src/anomalyzer/relationships.py","line":166,"excerpt":"source_indexes=source_indexes, source_timestamps=source_timestamps,"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":76,"excerpt":"(lambda d: d[\"timestamps\"].__setitem__(2, \"2026-01-01T02:30:00Z\"), \"off\"),"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":124,"excerpt":"def test_sparse_phase_without_support_is_explicitly_unavailable():"},{"reference":"tests/test_sparse.py","line":181,"excerpt":"ds[\"timestamps\"][1] = \"2026-02-01T00:00:00Z\""},{"reference":"docs/sparse-data.md","line":30,"excerpt":"timestamps=[\"2026-01-01T00:00:00Z\", \"2026-01-01T01:00:00Z\","},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":73,"excerpt":"conversion reasons, counts, revision IDs and availability timestamps. Converted"},{"reference":"docs/sparse-data.md","line":87,"excerpt":"otherwise bounds come from the supplied as-of-available timestamp range."},{"reference":"docs/sparse-data.md","line":105,"excerpt":"Duplicate timestamps normally fail. Legacy schema-1.1 relationship preparation"},{"reference":"examples/sparse_examples.py","line":59,"excerpt":"daily = request(dataset([1.0]*60, data_policy={\"measurement\":\"additive_volume\"}), recipe=\"multi-resolution-v1\","},{"reference":"examples/sparse_examples.py","line":95,"excerpt":"daily = next(item for item in outputs[\"unfinished_daily\"].dataset_results if item.resolution == \"completed_period\")"},{"reference":"examples/render_readme_progression.py","line":686,"excerpt":"def figure_multi_resolution():"},{"reference":"examples/render_readme_progression.py","line":699,"excerpt":"\"config\": {\"recipe\": \"multi-resolution-v1\", \"aggregate_function\": \"sum\"},"},{"reference":"examples/render_readme_progression.py","line":710,"excerpt":"assert partial[\"included_in_completed_period_view\"] is False"},{"reference":"examples/render_readme_progression.py","line":767,"excerpt":"complete_count = result.data_quality[\"views\"][\"completed_period\"][\"observation_count\"]"},{"reference":"examples/render_readme_progression.py","line":884,"excerpt":"header(svg, \"14\", \"Choose what an empty period means\", \"Identical source timestamps and values; only the declared empty-data policy changes.\")"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":51,"excerpt":"def valid_reset_points(cls, values):"},{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":71,"excerpt":"def numeric_values(cls, data):"},{"reference":"src/anomalyzer/contracts.py","line":85,"excerpt":"if self.period_statuses is not None and len(self.period_statuses) != len(self.values):"},{"reference":"src/anomalyzer/contracts.py","line":86,"excerpt":"raise ValueError(\"period_statuses and values must have equal length\")"},{"reference":"src/anomalyzer/sparse.py","line":3,"excerpt":"Calendar coordinates remain dense only inside an explicitly bounded input grid."},{"reference":"src/anomalyzer/sparse.py","line":34,"excerpt":"if ds.data_policy or ds.coverage is not None or ds.available_at is not None or ds.revisions or any(v is None for v in ds.values) or \"incomplete\" in (ds.period_statuses or []):"},{"reference":"src/anomalyzer/sparse.py","line":42,"excerpt":"\"\"\"Return the explicit grid, eligible model values and inspectable provenance.\"\"\""},{"reference":"src/anomalyzer/sparse.py","line":48,"excerpt":"statuses = ds.period_statuses or [\"complete\"] * len(ds.values)"},{"reference":"src/anomalyzer/sparse.py","line":54,"excerpt":"\"complete\" if value is not None else \"empty\" if policy.absence_means_zero else \"unknown\")"},{"reference":"src/anomalyzer/sparse_views.py","line":72,"excerpt":"measurements = quality[\"rule_values\"]"},{"reference":"src/anomalyzer/sparse_views.py","line":75,"excerpt":"if policy.observation_basis != \"calendar\":"},{"reference":"src/anomalyzer/sparse_views.py","line":82,"excerpt":"if policy.measurement != \"additive_volume\" or cfg.aggregate_function != \"sum\":"},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":98,"excerpt":"daily_times, daily_values, coverage, records, available_at = [], [], [], [], []"},{"reference":"src/anomalyzer/orchestrator.py","line":28,"excerpt":"if not any(row.get(\"available_at\") and (row.get(\"value\") is not None or row.get(\"coverage\") == \"excluded\") for row in samples):"},{"reference":"src/anomalyzer/orchestrator.py","line":63,"excerpt":"unavailable values. Dropping only that leading run lets the ordinary engine"},{"reference":"src/anomalyzer/orchestrator.py","line":127,"excerpt":"prepared_points += len(source.values) + len(dataset.revisions)"},{"reference":"src/anomalyzer/orchestrator.py","line":139,"excerpt":"source.quality.get(\"rule_values\", source.values)[:source.quality.get(\"grid_count\", source.quality[\"observation_count\"])], dataset.rules,"},{"reference":"src/anomalyzer/orchestrator.py","line":191,"excerpt":"[dict(value=p.value, available_at=p.latest_source_cutoff if any(p.source_available_at.values()) else None) for p in lineage])"},{"reference":"src/anomalyzer/relationships.py","line":36,"excerpt":"sources = [replace(source, values=source.quality.get(\"rule_values\", source.values)) for source in sources]"},{"reference":"src/anomalyzer/relationships.py","line":58,"excerpt":"x = predictor.values[index - relationship.lag]"},{"reference":"src/anomalyzer/relationships.py","line":98,"excerpt":"value, unavailable = None, \"zero_denominator\""},{"reference":"src/anomalyzer/relationships.py","line":131,"excerpt":"y, x = response.values[index], predictor.values[predictor_index]"},{"reference":"src/anomalyzer/relationships.py","line":141,"excerpt":"source_values = [source.values[source_index]"},{"reference":"src/anomalyzer/early.py","line":58,"excerpt":"for index, (observed_at, value) in enumerate(zip(times, values)):"},{"reference":"src/anomalyzer/early.py","line":137,"excerpt":"criterion_met and policy.allow_explicit_rules)"},{"reference":"tests/test_sparse.py","line":45,"excerpt":"def test_zero_is_only_authorized_for_complete_empty_additive_volume():"},{"reference":"tests/test_sparse.py","line":46,"excerpt":"request = body([2, None, None, 0, 9], data_policy={\"missing\": \"zero_for_empty\", \"measurement\": \"additive_volume\"},"},{"reference":"tests/test_sparse.py","line":51,"excerpt":"assert [r[\"original_value\"] for r in rows] == [2, None, None, 0, 9]"},{"reference":"tests/test_sparse.py","line":52,"excerpt":"assert rows[1][\"reason\"] == \"confirmed_empty_additive_volume\""},{"reference":"tests/test_sparse.py","line":58,"excerpt":"def test_global_collection_contract_authorizes_omitted_slots_but_not_explicit_unknown():"},{"reference":"docs/sparse-data.md","line":4,"excerpt":"Schema 1.1 therefore preserves missing values by default. An expectation can be"},{"reference":"docs/sparse-data.md","line":33,"excerpt":"data_policy=DataPolicy(),  # preserve missing; hour 2 is also unknown"},{"reference":"docs/sparse-data.md","line":41,"excerpt":"| Numeric value, including zero, `coverage=\"complete\"` | A measurement eligible for rules and model support. Explicit zero remains zero. |"},{"reference":"docs/sparse-data.md","line":43,"excerpt":"| `coverage=\"empty\"` | Caller confirms a finalized period with no reported volume. Preserve as missing unless an additive zero policy explicitly authorizes conversion. |"},{"reference":"docs/sparse-data.md","line":44,"excerpt":"| `coverage=\"unknown\"` | Collection/exposure is unknown; never filled even under a global empty-volume contract. |"},{"reference":"examples/sparse_examples.py","line":22,"excerpt":"preserve = request(dataset(values, coverage=[\"empty\" if v is None else \"complete\" for v in values],"},{"reference":"examples/sparse_examples.py","line":23,"excerpt":"data_policy={\"measurement\": \"additive_volume\"}))"},{"reference":"examples/sparse_examples.py","line":25,"excerpt":"zero[\"datasets\"][0][\"data_policy\"][\"missing\"] = \"zero_for_empty\""},{"reference":"examples/sparse_examples.py","line":26,"excerpt":"result.update(preserve=preserve, zero_empty=zero)"},{"reference":"examples/sparse_examples.py","line":28,"excerpt":"coverage=[\"complete\"]*2+[\"unknown\"]+[\"complete\"]*6+[\"unknown\"]+[\"complete\"]*3+[\"incomplete\"], data_policy={}))"},{"reference":"examples/render_readme_progression.py","line":135,"excerpt":"def chart(svg, values, overlays, x, y, width, height, *, y_bounds=None,"},{"reference":"examples/render_readme_progression.py","line":139,"excerpt":"all_values = [values] + [series for _, series, *_ in overlays]"},{"reference":"examples/render_readme_progression.py","line":140,"excerpt":"low, high = y_bounds or bounds(*all_values)"},{"reference":"examples/render_readme_progression.py","line":166,"excerpt":"if value is not None and (index == 0 or values[index-1] is None) and (index == count-1 or values[index+1] is None):"},{"reference":"examples/render_readme_progression.py","line":206,"excerpt":"values = [80 + 0.65 * index + [0, 0.25, -0.15, 0.10, -0.20][index % 5]"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":81,"excerpt":"elif self.frequency is None and getattr(getattr(self, \"data_policy\", None), \"observation_basis\", None) != \"event_index\":"},{"reference":"src/anomalyzer/contracts.py","line":242,"excerpt":"missing: Literal[\"preserve\", \"zero_for_empty\"] = \"preserve\""},{"reference":"src/anomalyzer/contracts.py","line":245,"excerpt":"observation_basis: Literal[\"calendar\", \"event_index\"] = \"calendar\""},{"reference":"src/anomalyzer/contracts.py","line":258,"excerpt":"if self.observation_basis == \"event_index\" and ("},{"reference":"src/anomalyzer/sparse.py","line":4,"excerpt":"Event coordinates retain actual timestamps and represent comparable events, not time."},{"reference":"src/anomalyzer/sparse.py","line":49,"excerpt":"for index, (text, value, status) in enumerate(zip(ds.timestamps, ds.values, statuses)):"},{"reference":"src/anomalyzer/sparse.py","line":57,"excerpt":"available = timestamp(ds.available_at[index], ds.timezone) if ds.available_at and ds.available_at[index] else None"},{"reference":"src/anomalyzer/sparse.py","line":63,"excerpt":"if cfg.duplicate_policy == \"error\" or policy.observation_basis == \"event_index\":"},{"reference":"src/anomalyzer/sparse.py","line":64,"excerpt":"raise ValueError(\"duplicate timestamps require declared duplicate_policy; event_index needs unique event timestamps\")"},{"reference":"src/anomalyzer/sparse_views.py","line":76,"excerpt":"unsupported = \"completed-day aggregation is unavailable on event_index\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/sparse_views.py","line":128,"excerpt":"policy_zero_children=sum(quality[\"samples\"][i][\"reason\"] == \"confirmed_empty_additive_volume\" for i in indexes),"},{"reference":"src/anomalyzer/sparse_views.py","line":129,"excerpt":"source_input_indexes=[raw for i in indexes for raw in quality[\"samples\"][i][\"source_indexes\"]],"},{"reference":"src/anomalyzer/sparse_views.py","line":134,"excerpt":"timestamps=daily_times, values=daily_values, frequency=\"1d\", coverage=coverage, available_at=available_at,"},{"reference":"src/anomalyzer/orchestrator.py","line":37,"excerpt":"assessment.baseline[\"known_input_availability_cutoff\"] = cutoffs[assessment.index]"},{"reference":"src/anomalyzer/orchestrator.py","line":38,"excerpt":"assessment.assumptions.append(\"Availability cutoff is conservative across known prior inputs; unknown arrival history cannot establish delivery-time detection performance.\")"},{"reference":"src/anomalyzer/orchestrator.py","line":166,"excerpt":"if any(definitions[name].timestamps is None for name in _source_ids(relationship)):"},{"reference":"src/anomalyzer/orchestrator.py","line":178,"excerpt":"relationship.id, derived.timestamps or [], derived.values,"},{"reference":"src/anomalyzer/orchestrator.py","line":180,"excerpt":"reset_positions=active_resets(request.config.reset_points, derived.timestamps or [], derived.timezone, request.context.as_of),"},{"reference":"src/anomalyzer/relationships.py","line":40,"excerpt":"raise ValueError(\"relationship timestamps cannot be positional\")"},{"reference":"src/anomalyzer/relationships.py","line":57,"excerpt":"for index in range(relationship.lag, coefficient_cutoff + 1):"},{"reference":"src/anomalyzer/relationships.py","line":58,"excerpt":"x = predictor.values[index - relationship.lag]"},{"reference":"src/anomalyzer/relationships.py","line":96,"excerpt":"if relationship.zero_denominator == \"error\":"},{"reference":"src/anomalyzer/relationships.py","line":97,"excerpt":"raise ValueError(f\"zero denominator at aligned index {index}\")"},{"reference":"src/anomalyzer/early.py","line":140,"excerpt":"unavailable_rules = [dict(id=rule.id, reason=\"no_comparable_prior\" if previous is None else \"relative_change_from_zero\")"},{"reference":"src/anomalyzer/early.py","line":147,"excerpt":"target_id=target_id, index=index, timestamp=observed_at,"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":68,"excerpt":"{\"observation_basis\": \"event_index\", \"grid_start\": \"2026-01-01T00:00:00Z\"},"},{"reference":"tests/test_sparse.py","line":76,"excerpt":"(lambda d: d[\"timestamps\"].__setitem__(2, \"2026-01-01T02:30:00Z\"), \"off\"),"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":129,"excerpt":"assert any(row[\"reason\"] == \"no_training_support_for_calendar_phase\" for row in result.methods[0].diagnostics[\"unavailable\"])"},{"reference":"docs/sparse-data.md","line":30,"excerpt":"timestamps=[\"2026-01-01T00:00:00Z\", \"2026-01-01T01:00:00Z\","},{"reference":"docs/sparse-data.md","line":105,"excerpt":"Duplicate timestamps normally fail. Legacy schema-1.1 relationship preparation"},{"reference":"docs/sparse-data.md","line":110,"excerpt":"rejected. Event-index timestamps must be unique; ties are refused rather than"},{"reference":"docs/sparse-data.md","line":141,"excerpt":"original calendar indexes and timestamps. Post-gap learning requires an explicit"},{"reference":"docs/sparse-data.md","line":155,"excerpt":"## Comparable event-index observations"},{"reference":"examples/sparse_examples.py","line":26,"excerpt":"result.update(preserve=preserve, zero_empty=zero)"},{"reference":"examples/sparse_examples.py","line":36,"excerpt":"events = dataset(costs, units=\"dollars_per_call\", data_policy={\"observation_basis\": \"event_index\"})"},{"reference":"examples/sparse_examples.py","line":38,"excerpt":"events[\"timestamps\"] = [(datetime(2026,1,1,tzinfo=timezone.utc)+timedelta(hours=h)).isoformat() for h in hours]"},{"reference":"examples/sparse_examples.py","line":57,"excerpt":"relation[\"relationships\"] = [{\"id\":\"unit-cost\",\"kind\":\"ratio\",\"numerator\":\"series\",\"denominator\":\"calls\",\"units\":\"dollars_per_call\"}]"},{"reference":"examples/sparse_examples.py","line":81,"excerpt":"preserve, zero = [outputs[n].dataset_results[0] for n in (\"preserve\", \"zero_empty\")]"},{"reference":"examples/render_readme_progression.py","line":654,"excerpt":"svg.text(430, 467, \"Sample index (zero-based)\", 15, MUTED, anchor=\"middle\")"},{"reference":"examples/render_readme_progression.py","line":697,"excerpt":"\"timestamps\": [(origin + timedelta(hours=index)).isoformat()"},{"reference":"examples/render_readme_progression.py","line":884,"excerpt":"header(svg, \"14\", \"Choose what an empty period means\", \"Identical source timestamps and values; only the declared empty-data policy changes.\")"},{"reference":"examples/render_readme_progression.py","line":887,"excerpt":"shared_high = max(v for v in observed(\"preserve\")+expectation(\"preserve\")+observed(\"zero_empty\")+expectation(\"zero_empty\") if v is not None)*1.12"},{"reference":"examples/render_readme_progression.py","line":897,"excerpt":"header(svg, \"15\", \"Keep zero, unknown and unfinished distinct\", \"Blue zeros are actual measurements; gaps stay gaps; the orange open-period value is excluded.\")"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Assess sparse source data","order":30}},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","status":"implemented","area":"data","path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","health":"review","coverage":"partial","summary":"Explicitly incomplete observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"Schema-1.0 incomplete observations form a trailing suffix; schema-1.1 source policies preserve internal and trailing incomplete positions as unavailable.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","order":10}},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","status":"implemented","area":"data","path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","health":"review","coverage":"partial","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"Calendar datasets provide timestamps and frequency; positional datasets provide neither. Explicit schema-1.1 event-index sources retain timestamps without a calendar frequency.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","order":20}},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","status":"implemented","area":"execution","path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","health":"review","coverage":"partial","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","order":50}},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","health":"review","coverage":"partial","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","order":50}},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","health":"review","coverage":"partial","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Readiness status remains visible, including when it was not assessed.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","order":60}},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","health":"review","coverage":"partial","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Analysis tool calls return structured content together with compact text.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","order":10}},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","order":20}},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","health":"review","coverage":"partial","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","order":10}},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","health":"review","coverage":"partial","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","order":10}},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","health":"review","coverage":"partial","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","order":20}},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","status":"implemented","area":"investigation","path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","health":"review","coverage":"partial","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","order":10}},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","status":"implemented","area":"investigation","path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","order":40}},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","status":"implemented","area":"investigation","path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","health":"review","coverage":"partial","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Default policy leaves action and notification eligibility disabled.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","order":60}},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","status":"implemented","area":"investigation","path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","health":"review","coverage":"partial","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","order":30}},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","health":"review","coverage":"partial","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","order":20}},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","health":"review","coverage":"partial","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","order":30}},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","status":"implemented","area":"relationships","path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","health":"review","coverage":"partial","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"Difference relationships require identical declared source units and matching output units.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","order":10}},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","status":"implemented","area":"relationships","path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","health":"review","coverage":"partial","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed fixed UTC 24-hour additive aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Run semantic review, save review evidence, or ignore accepted findings.","topFindings":[{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":19,"excerpt":"def view_cases(datasets, relationships, request):"},{"reference":"src/anomalyzer/sparse_views.py","line":22,"excerpt":"cases = compose_cases([d for d in datasets if d.resolution == resolution],"},{"reference":"src/anomalyzer/sparse_views.py","line":146,"excerpt":"native_sources = [d.model_copy(update={\"data_policy\": d.data_policy or DataPolicy()}) for d in request.datasets]"},{"reference":"src/anomalyzer/sparse_views.py","line":149,"excerpt":"native_request = request.model_copy(update={\"datasets\": native_sources, \"config\": native_config})"},{"reference":"src/anomalyzer/sparse_views.py","line":169,"excerpt":"daily_request = request.model_copy(update={\"datasets\": sources, \"config\": daily_config, \"relationships\": daily_relationships})"},{"reference":"src/anomalyzer/agent.py","line":14,"excerpt":"\"\"\"Use legacy semantics for a Dataset; explicit V11 fields select source policies.\"\"\""},{"reference":"src/anomalyzer/agent.py","line":20,"excerpt":"name in dataset for name in (\"data_policy\", \"coverage\", \"available_at\", \"revisions\", \"rules\", \"completed_period_rules\")):"},{"reference":"src/anomalyzer/agent.py","line":23,"excerpt":"return analyze({\"schema_version\": \"1.1\", \"datasets\": [validated.model_dump(mode=\"json\")],"},{"reference":"src/anomalyzer/agent.py","line":29,"excerpt":"return analyze({\"datasets\": [validated.model_dump(mode=\"json\")],"},{"reference":"tests/test_sparse.py","line":18,"excerpt":"return {\"schema_version\": \"1.1\", \"datasets\": [{\"id\": \"series\", \"entity\": {\"customer\": \"test\"},"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":88,"excerpt":"def test_declared_duplicate_aggregation_preserves_provenance(policy, value):"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":118,"excerpt":"changed[\"datasets\"][0][\"values\"][-10:] = [999]*10"},{"reference":"docs/sparse-data.md","line":35,"excerpt":"result = analyze({\"schema_version\": \"1.1\", \"datasets\": [source.model_dump(mode=\"json\")]})"},{"reference":"docs/sparse-data.md","line":136,"excerpt":"incomplete, excluded or reset boundaries, regardless of point-reference policy."},{"reference":"docs/sparse-data.md","line":142,"excerpt":"reset or caller-selected `reference_gaps=\"restart\"`; the default does not silently"},{"reference":"docs/sparse-data.md","line":172,"excerpt":"basis. A caller-declared maximum time separation can limit change comparisons."},{"reference":"docs/sparse-data.md","line":218,"excerpt":"Daily lagged, normalized-residual and joint-condition definitions remain"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":29,"excerpt":"case.non_claims.append(\"Native and completed-day views reuse the same observations; they are not independent confirmation.\")"},{"reference":"src/anomalyzer/sparse_views.py","line":70,"excerpt":"\"\"\"Sum only wholly observed finalized additive children. Never average ratios.\"\"\""},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":175,"excerpt":"applicability=\"completed-day view supports ratio-of-sums and differences of compatible additive sums only\","},{"reference":"src/anomalyzer/relationships.py","line":90,"excerpt":"used = [(numerator, index), (denominator, index)]"},{"reference":"src/anomalyzer/relationships.py","line":173,"excerpt":"\"complete\" if source.values[i] is not None else \"unknown_or_incomplete\" for source, i in used}"},{"reference":"tests/test_sparse.py","line":231,"excerpt":"request[\"relationships\"] = [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"series\", \"denominator\": \"calls\", \"units\": \"dollars_per_call\"}]"},{"reference":"tests/test_sparse.py","line":260,"excerpt":"def test_daily_rules_are_not_silently_reused_from_hourly_rules():"},{"reference":"tests/test_sparse.py","line":279,"excerpt":"def test_adaptive_gap_does_not_silently_train_a_post_gap_episode():"},{"reference":"tests/test_sparse.py","line":322,"excerpt":"def test_restart_policy_cannot_reuse_a_prior_regime_reference():"},{"reference":"tests/test_sparse.py","line":431,"excerpt":"assert row[\"complete\"] and row[\"aggregate_value\"] == 24 and not row[\"model_eligible\"]"},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":120,"excerpt":"It then collects `calibration_size` eligible forecast residuals, not elapsed slots."},{"reference":"docs/sparse-data.md","line":142,"excerpt":"reset or caller-selected `reference_gaps=\"restart\"`; the default does not silently"},{"reference":"docs/sparse-data.md","line":213,"excerpt":"partial sums are progress metadata, never a completed total. Empty days are not"},{"reference":"docs/sparse-data.md","line":216,"excerpt":"Daily ratios divide summed compatible numerator and denominator quantities;"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":19,"excerpt":"def view_cases(datasets, relationships, request):"},{"reference":"src/anomalyzer/sparse_views.py","line":76,"excerpt":"unsupported = \"completed-day aggregation is unavailable on event_index\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/sparse_views.py","line":129,"excerpt":"source_input_indexes=[raw for i in indexes for raw in quality[\"samples\"][i][\"source_indexes\"]],"},{"reference":"src/anomalyzer/sparse_views.py","line":130,"excerpt":"source_grid_indexes=indexes, observed_sum=(math.fsum(measurements[i] for i in indexes if measurements[i] is not None)"},{"reference":"src/anomalyzer/relationships.py","line":7,"excerpt":"from .alignment import PreparedDataset, exact_sources"},{"reference":"src/anomalyzer/relationships.py","line":19,"excerpt":"def _source_ids(relationship: Relationship) -> list[str]:"},{"reference":"src/anomalyzer/relationships.py","line":33,"excerpt":"ids = _source_ids(relationship)"},{"reference":"src/anomalyzer/relationships.py","line":34,"excerpt":"sources = exact_sources(ids, prepared)"},{"reference":"src/anomalyzer/relationships.py","line":36,"excerpt":"sources = [replace(source, values=source.quality.get(\"rule_values\", source.values)) for source in sources]"},{"reference":"src/anomalyzer/agent.py","line":34,"excerpt":"def analyze_relationships(request: RequestV11 | dict) -> ResultV11:"},{"reference":"tests/test_sparse.py","line":10,"excerpt":"from anomalyzer import analyze, analyze_series, analyze_relationships, get_case, replay_policy"},{"reference":"tests/test_sparse.py","line":41,"excerpt":"assert right.data_quality[\"samples\"][8][\"source_index\"] == -1"},{"reference":"tests/test_sparse.py","line":51,"excerpt":"assert [r[\"original_value\"] for r in rows] == [2, None, None, 0, 9]"},{"reference":"tests/test_sparse.py","line":61,"excerpt":"assert [r[\"value\"] for r in source(request).data_quality[\"samples\"]] == [1, 0, 0, 2]"},{"reference":"tests/test_sparse.py","line":63,"excerpt":"assert [r[\"value\"] for r in source(request).data_quality[\"samples\"]] == [1, None, 0, 2]"},{"reference":"docs/sparse-data.md","line":21,"excerpt":"works through `analyze`, `analyze_relationships`, CLI JSON and read-only MCP."},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":72,"excerpt":"Preparation reports original values, model values, coverage, source indexes,"},{"reference":"docs/sparse-data.md","line":76,"excerpt":"For policy-prepared sources, `grid_count` counts positions, `observation_count`"},{"reference":"docs/sparse-data.md","line":106,"excerpt":"retains its unique-source requirement. With an explicit `data_policy`, complete,"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Related-metric analysis","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","order":30}},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","status":"planned","area":"analysis","path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","health":"planned","coverage":"uncovered","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","order":10}},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","status":"planned","area":"analysis","path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","health":"planned","coverage":"uncovered","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","order":20}},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","status":"planned","area":"data","path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","order":10}},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","status":"planned","area":"data","path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","health":"planned","coverage":"uncovered","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","order":10}},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","status":"planned","area":"data","path":".capabilities/data/quality/report-source-freshness.capability.yaml","health":"planned","coverage":"uncovered","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","order":20}},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","status":"planned","area":"data","path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","order":30}},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","status":"planned","area":"evaluation","path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","order":30}},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","status":"planned","area":"evaluation","path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","health":"planned","coverage":"uncovered","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","order":20}},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","status":"planned","area":"evaluation","path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","order":10}},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","status":"planned","area":"execution","path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","order":20}},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","status":"planned","area":"execution","path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","order":10}},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","status":"planned","area":"interfaces","path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","order":20}},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","status":"planned","area":"investigation","path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","order":20}},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","order":10}},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","health":"planned","coverage":"uncovered","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","order":30}},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","status":"planned","area":"investigation","path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","order":30}},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","status":"planned","area":"relationships","path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","order":30}}],"byStoryMap":{"unassigned":[],"releases":[{"release":"Anomaly detection Skill","capabilities":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/guide-anomaly-investigation.capability.yaml","health":"review","coverage":"partial","summary":"The reusable anomaly-detection Skill guides an agent through local Anomalyzer analysis and a concise explanation grounded in structured evidence.","intent":"Make the analysis usable through an agent while keeping data handling and interpretation limits explicit.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The default response instructions lead with changes and observed-versus-expected values, then explain readiness, evidence maturity, overlapping findings, and material limitations without inventing confidence percentages or causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill describes MCP analyze, list_findings, and get_evidence as an optional evidence workflow and makes session lifetime, read-only access, and caller-owned notification and action delivery explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/guide-anomaly-investigation","title":"Ask an agent to explain unusual changes in my metrics","criterion":"The Skill directs analysis through the installed Anomalyzer CLI, with executable fallback paths and structured JSON output, rather than a newly invented detector.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/guide-anomaly-investigation --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","order":10}},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","status":"implemented","area":"interfaces","path":".capabilities/interfaces/skill/visualize-analysis-on-request.capability.yaml","health":"review","coverage":"partial","summary":"The anomaly-detection Skill offers an explicitly requested quick chart or a detailed figure specification after structured analysis completes.","intent":"Make findings easier to inspect visually while preserving their evidence meaning and the user's requested level of detail.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"Detailed figure instructions preserve source indices, provisional-versus-calibrated evidence, pattern detection timing, and incomplete or failed run status, without presenting scores as probabilities or flags as proven causes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The dependency-free quick chart fallback renders one resolved dataset's observed values and available expectations, marks only triggered evidence, and labels training, calibration, units, and reviewed reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/skill/visualize-analysis-on-request","title":"See the evidence in a chart when I ask for one","criterion":"The fallback rejects unsupported multi-dataset results and protects an existing output file unless overwrite is explicitly requested.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/skill/visualize-analysis-on-request --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Anomaly detection Skill","backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","order":20}}],"deliveryStrategy":{"release":"Anomaly detection Skill","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/skill/guide-anomaly-investigation"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/skill/guide-anomaly-investigation needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Ask an agent to investigate a change"],"missingBackbones":[],"stepCoverageGaps":["Ask an agent to investigate a change > Inspect a requested chart"],"rationale":"Starts with the earliest mapped step in Ask an agent to investigate a change."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/skill/visualize-analysis-on-request"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/skill/visualize-analysis-on-request needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Ask an agent to investigate a change"],"missingBackbones":[],"stepCoverageGaps":["Ask an agent to investigate a change > Get an evidence-grounded explanation"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/skill/visualize-analysis-on-request."}]},"presentation":{"outcome":"Anomaly detection Skill helps teams move from Get an evidence-grounded explanation to Inspect a requested chart across 1 backbone activity.","narrativePath":[{"backbone":"Ask an agent to investigate a change","step":"Get an evidence-grounded explanation","capabilityIds":["interfaces/skill/guide-anomaly-investigation"],"health":"review"},{"backbone":"Ask an agent to investigate a change","step":"Inspect a requested chart","capabilityIds":["interfaces/skill/visualize-analysis-on-request"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Ask an agent to investigate a change > Get an evidence-grounded explanation","message":"Ask an agent to investigate a change > Get an evidence-grounded explanation has needs-review coverage."},{"kind":"weak","label":"Ask an agent to investigate a change > Inspect a requested chart","message":"Ask an agent to investigate a change > Inspect a requested chart has needs-review coverage."}]}},{"release":"Basic CLI anomaly detection","capabilities":[{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/report-point-anomaly-episodes.capability.yaml","health":"review","coverage":"partial","summary":"Calibrated point-threshold crossings become inspectable observations, with adjacent crossings grouped into episodes.","intent":"Locate unusual values and keep isolated events distinct from sustained runs.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"A sufficiently extreme calibrated sample records a point trigger and the expectation it departed from.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"Adjacent point departures form a consecutive-run pattern with the contributing sample positions.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/report-point-anomaly-episodes","title":"Find unusual spikes and drops","criterion":"Repeated runs of the same fixed request preserve numerical evidence; future samples do not change earlier scoring for declared settings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/report-point-anomaly-episodes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Find spikes and drops","order":40}},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/calibrate-departures.capability.yaml","health":"review","coverage":"partial","summary":"Sample evidence distinguishes reference-only, provisional, and calibrated comparisons before permitting statistical point triggers.","intent":"Keep useful early evidence while making its level of support visible.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Completed calibration provides a frozen residual center and sample scale for later standardized residuals.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Evidence records the observed value, expectation, calibration support, and numerical criterion responsible for a trigger.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/calibrate-departures","title":"Understand how much history supports a finding","criterion":"Provisional residual scores can be reported but do not trigger point-anomaly episodes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/calibrate-departures --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Establish evidence support","order":30}},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","status":"implemented","area":"data","path":".capabilities/data/inputs/accept-analysis-requests.capability.yaml","health":"review","coverage":"partial","summary":"Callers can submit ordered numeric values or versioned requests containing datasets, settings, context, and declared relationships.","intent":"Make the accepted data contract clear before any conclusions are drawn from the numbers.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Ordered value lists work with separate settings, while schema 1.0 requests describe one dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Published schema 1.0 and 1.1 request and result schemas match the public Python contracts.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/inputs/accept-analysis-requests","title":"Bring my metrics into an analysis","criterion":"Schema 1.1 datasets declare a nonempty entity scope and support explicit relationships between named datasets.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/inputs/accept-analysis-requests --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Supply a numeric series","order":10}},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","status":"implemented","area":"data","path":".capabilities/data/quality/prepare-timestamped-series.capability.yaml","health":"review","coverage":"partial","summary":"Timestamped observations are normalized to UTC, ordered, and checked against their declared cadence before analysis.","intent":"Compare like periods and make preparation decisions visible to the caller.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"An as_of cutoff excludes later observations, and cadence is checked as a fixed elapsed duration.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"Calendar datasets provide timestamps and frequency; positional datasets provide neither. Explicit schema-1.1 event-index sources retain timestamps without a calendar frequency.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/prepare-timestamped-series","title":"Analyze observations in the right time order","criterion":"Duplicate timestamps are rejected unless an aggregation policy is supplied; duplicate counts and transformations remain visible.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/prepare-timestamped-series --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Prepare the comparison","order":20}},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/analyze-files-and-streams.capability.yaml","health":"review","coverage":"partial","summary":"The CLI analyzes CSV, JSON requests, and bare JSON values from files or stdin, with configurable settings and protected output files.","intent":"Make local analysis usable in scripts and ad hoc investigation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"CSV supports positional or timestamped selected values, treats empty cells as missing, and rejects varying unmapped entity columns.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Equivalent file and stdin JSON inputs produce equivalent request fingerprints and analysis semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/analyze-files-and-streams","title":"Analyze local files from the command line","criterion":"Output files require explicit overwrite permission, and invalid input is distinguished from execution failure by exit status.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/analyze-files-and-streams --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Run from files or stdin","order":50}},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","status":"implemented","area":"interfaces","path":".capabilities/interfaces/cli/explain-findings-in-readable-output.capability.yaml","health":"review","coverage":"partial","summary":"Readable CLI output distinguishes point episodes, pattern findings, cases, explicit rule violations, and reference readiness.","intent":"Let a person understand what changed and which evidence supports it without parsing the full JSON result.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Explicit numerical rules are named with their boundaries instead of appearing as statistical anomaly claims.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Point-anomaly episode counts and residual-pattern findings are labeled separately.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/cli/explain-findings-in-readable-output","title":"Understand findings in a readable summary","criterion":"Readiness status remains visible, including when it was not assessed.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/cli/explain-findings-in-readable-output --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Basic CLI anomaly detection","backbone":"Analyze a local series","step":"Read the findings","order":60}}],"deliveryStrategy":{"release":"Basic CLI anomaly detection","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/inputs/accept-analysis-requests"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/inputs/accept-analysis-requests needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Analyze a local series"],"missingBackbones":[],"stepCoverageGaps":["Analyze a local series > Establish evidence support","Analyze a local series > Find spikes and drops","Analyze a local series > Prepare the comparison","Analyze a local series > Read the findings","Analyze a local series > Run from files or stdin"],"rationale":"Starts with the earliest mapped step in Analyze a local series."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["data/quality/prepare-timestamped-series","analysis/evidence/calibrate-departures","analysis/detection/report-point-anomaly-episodes","interfaces/cli/analyze-files-and-streams","interfaces/cli/explain-findings-in-readable-output"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"analysis/detection/report-point-anomaly-episodes needs review; analysis/evidence/calibrate-departures needs review; data/quality/prepare-timestamped-series needs review; interfaces/cli/analyze-files-and-streams needs review; interfaces/cli/explain-findings-in-readable-output needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Analyze a local series"],"missingBackbones":[],"stepCoverageGaps":["Analyze a local series > Supply a numeric series"],"rationale":"Prioritizes capabilities with delivery risk signals: data/quality/prepare-timestamped-series, analysis/evidence/calibrate-departures, analysis/detection/report-point-anomaly-episodes, interfaces/cli/analyze-files-and-streams, interfaces/cli/explain-findings-in-readable-output."}]},"presentation":{"outcome":"Basic CLI anomaly detection helps teams move from Supply a numeric series to Read the findings across 1 backbone activity.","narrativePath":[{"backbone":"Analyze a local series","step":"Supply a numeric series","capabilityIds":["data/inputs/accept-analysis-requests"],"health":"review"},{"backbone":"Analyze a local series","step":"Prepare the comparison","capabilityIds":["data/quality/prepare-timestamped-series"],"health":"review"},{"backbone":"Analyze a local series","step":"Establish evidence support","capabilityIds":["analysis/evidence/calibrate-departures"],"health":"review"},{"backbone":"Analyze a local series","step":"Find spikes and drops","capabilityIds":["analysis/detection/report-point-anomaly-episodes"],"health":"review"},{"backbone":"Analyze a local series","step":"Run from files or stdin","capabilityIds":["interfaces/cli/analyze-files-and-streams"],"health":"review"},{"backbone":"Analyze a local series","step":"Read the findings","capabilityIds":["interfaces/cli/explain-findings-in-readable-output"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Analyze a local series > Establish evidence support","message":"Analyze a local series > Establish evidence support has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Find spikes and drops","message":"Analyze a local series > Find spikes and drops has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Prepare the comparison","message":"Analyze a local series > Prepare the comparison has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Read the findings","message":"Analyze a local series > Read the findings has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Run from files or stdin","message":"Analyze a local series > Run from files or stdin has needs-review coverage."},{"kind":"weak","label":"Analyze a local series > Supply a numeric series","message":"Analyze a local series > Supply a numeric series has needs-review coverage."}]}},{"release":"Evidence and review policy","capabilities":[{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","status":"implemented","area":"analysis","path":".capabilities/analysis/evidence/assess-reference-readiness.capability.yaml","health":"review","coverage":"partial","summary":"Readiness diagnostics report supported, caution, or not_assessed based on the training pattern and calibration evidence.","intent":"Expose weak expectations even when no anomaly is detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Insufficient calibration support yields not_assessed, including when the run otherwise completes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Later evaluation changes do not rewrite the segment's training-and-calibration readiness assessment.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/evidence/assess-reference-readiness","title":"Know when the expected baseline is weak","criterion":"Repeating seasonal structure with stable calibration can yield supported readiness.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/evidence/assess-reference-readiness --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","order":20}},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","status":"implemented","area":"execution","path":".capabilities/execution/resilience/retain-evidence-under-failure-and-limits.capability.yaml","health":"review","coverage":"partial","summary":"Analysis applies cooperative resource limits and exposes partial or failed results without discarding valid earlier or sibling evidence.","intent":"Make constrained or failed runs understandable instead of silently returning incomplete conclusions.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"A failed dataset or relationship leaves healthy sibling results available and marks the overall run partial.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"A method failure preserves evidence completed before the failure and reports the error.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"execution/resilience/retain-evidence-under-failure-and-limits","title":"Keep useful evidence when analysis cannot finish","criterion":"CLI byte limits and execution failures produce distinct error exits.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify execution/resilience/retain-evidence-under-failure-and-limits --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","order":50}},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","status":"implemented","area":"investigation","path":".capabilities/investigation/assessments/describe-early-changes.capability.yaml","health":"review","coverage":"partial","summary":"Short histories yield observations, changes, and departure candidates with explicit comparison populations and limits.","intent":"Provide an inspectable lead for investigation without overstating statistical support.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Descriptive assessments use preceding observations and remain unchanged when future samples are appended.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Early candidates remain non-triggering and ineligible for action even under a permissive policy.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/assessments/describe-early-changes","title":"Spot early changes with a short history","criterion":"Missing values break the comparison history, and zero previous values do not create relative-change evidence.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/assessments/describe-early-changes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","order":10}},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","status":"implemented","area":"investigation","path":".capabilities/investigation/cases/group-related-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Qualifying assessments are grouped by event time and entity into cases with explanations, evidence references, and numerical severity.","intent":"Give investigators a coherent account of related evidence without counting derived metrics as independent confirmation.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"A derived relationship adds context without counting as an additional independent source for severity.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"Adaptive location-shift rules can create an early grouped case without a point trigger, using only evidence available at detection.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/cases/group-related-evidence","title":"Investigate related findings together","criterion":"Criteria met in multiple source datasets can raise numerical severity and identify the contributing sources.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/cases/group-related-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","order":40}},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","status":"implemented","area":"investigation","path":".capabilities/investigation/policy/replay-action-eligibility.capability.yaml","health":"review","coverage":"partial","summary":"Callers can apply a different explicit action policy to existing assessments and cases without recomputing numerical analysis.","intent":"Separate an observed criterion from a caller's decision to make it eligible for further action.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Default policy leaves action and notification eligibility disabled.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Explicit-rule and calibrated-departure eligibility are controlled independently.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/policy/replay-action-eligibility","title":"Change action eligibility without rerunning analysis","criterion":"Notification eligibility can be allowed for a met criterion without action eligibility, while early candidates remain ineligible.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/policy/replay-action-eligibility --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Review action eligibility","order":60}},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","status":"implemented","area":"investigation","path":".capabilities/investigation/rules/check-declared-numerical-rules.capability.yaml","health":"review","coverage":"partial","summary":"Acceptable ranges and maximum absolute or relative changes identify explicit criterion violations, including before calibration.","intent":"Let callers check clear numerical constraints without presenting them as learned anomaly thresholds.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Absolute and relative changes must exceed their declared maximum to violate a rule.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Relative-change checks require a nonzero previous value and do not bridge missing observations.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"investigation/rules/check-declared-numerical-rules","title":"Find values that break my stated limits","criterion":"Values outside an inclusive acceptable range are violations, while values on the boundary are accepted.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify investigation/rules/check-declared-numerical-rules --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Evidence and review policy","backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","order":30}}],"deliveryStrategy":{"release":"Evidence and review policy","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["investigation/assessments/describe-early-changes"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"investigation/assessments/describe-early-changes needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Decide how much to rely on a finding"],"missingBackbones":[],"stepCoverageGaps":["Decide how much to rely on a finding > Apply my stated limits","Decide how much to rely on a finding > Assess baseline quality","Decide how much to rely on a finding > Inspect incomplete execution","Decide how much to rely on a finding > Investigate related evidence","Decide how much to rely on a finding > Review action eligibility"],"rationale":"Starts with the earliest mapped step in Decide how much to rely on a finding."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["analysis/evidence/assess-reference-readiness","investigation/rules/check-declared-numerical-rules","investigation/cases/group-related-evidence","execution/resilience/retain-evidence-under-failure-and-limits","investigation/policy/replay-action-eligibility"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"analysis/evidence/assess-reference-readiness needs review; execution/resilience/retain-evidence-under-failure-and-limits needs review; investigation/cases/group-related-evidence needs review; investigation/policy/replay-action-eligibility needs review; investigation/rules/check-declared-numerical-rules needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Decide how much to rely on a finding"],"missingBackbones":[],"stepCoverageGaps":["Decide how much to rely on a finding > Inspect early evidence"],"rationale":"Prioritizes capabilities with delivery risk signals: analysis/evidence/assess-reference-readiness, investigation/rules/check-declared-numerical-rules, investigation/cases/group-related-evidence, execution/resilience/retain-evidence-under-failure-and-limits, investigation/policy/replay-action-eligibility."}]},"presentation":{"outcome":"Evidence and review policy helps teams move from Inspect early evidence to Review action eligibility across 1 backbone activity.","narrativePath":[{"backbone":"Decide how much to rely on a finding","step":"Inspect early evidence","capabilityIds":["investigation/assessments/describe-early-changes"],"health":"review"},{"backbone":"Decide how much to rely on a finding","step":"Assess baseline quality","capabilityIds":["analysis/evidence/assess-reference-readiness"],"health":"review"},{"backbone":"Decide how much to rely on a finding","step":"Apply my stated limits","capabilityIds":["investigation/rules/check-declared-numerical-rules"],"health":"review"},{"backbone":"Decide how much to rely on a finding","step":"Investigate related evidence","capabilityIds":["investigation/cases/group-related-evidence"],"health":"review"},{"backbone":"Decide how much to rely on a finding","step":"Inspect incomplete execution","capabilityIds":["execution/resilience/retain-evidence-under-failure-and-limits"],"health":"review"},{"backbone":"Decide how much to rely on a finding","step":"Review action eligibility","capabilityIds":["investigation/policy/replay-action-eligibility"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Decide how much to rely on a finding > Apply my stated limits","message":"Decide how much to rely on a finding > Apply my stated limits has needs-review coverage."},{"kind":"weak","label":"Decide how much to rely on a finding > Assess baseline quality","message":"Decide how much to rely on a finding > Assess baseline quality has needs-review coverage."},{"kind":"weak","label":"Decide how much to rely on a finding > Inspect early evidence","message":"Decide how much to rely on a finding > Inspect early evidence has needs-review coverage."},{"kind":"weak","label":"Decide how much to rely on a finding > Inspect incomplete execution","message":"Decide how much to rely on a finding > Inspect incomplete execution has needs-review coverage."},{"kind":"weak","label":"Decide how much to rely on a finding > Investigate related evidence","message":"Decide how much to rely on a finding > Investigate related evidence has needs-review coverage."},{"kind":"weak","label":"Decide how much to rely on a finding > Review action eligibility","message":"Decide how much to rely on a finding > Review action eligibility has needs-review coverage."}]}},{"release":"Future - Business-aware expectations","capabilities":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","status":"planned","area":"analysis","path":".capabilities/analysis/expectations/model-multiple-seasonal-cycles.capability.yaml","health":"planned","coverage":"uncovered","summary":"Build an inspectable expectation from a bounded set of caller-declared seasonal cycles instead of forcing one cycle to explain every repeating pattern.","intent":"Reduce departures caused by ordinary combinations such as hour-of-day and weekday behavior while retaining evidence of unusual changes.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated or manual verification checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/expectations/model-multiple-seasonal-cycles has no automated checks","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"},{"code":"missing-manual-review","message":"analysis/expectations/model-multiple-seasonal-cycles has no manual review guidance","capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\expectations\\model-multiple-seasonal-cycles.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Calibration remains separated from expectation fitting, and an isolated extreme sample retains evidence without silently contaminating every cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Callers explicitly declare supported seasonal cycles and their elapsed or calendar meaning; duplicate or incompatible cycle definitions are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."},{"capabilityId":"analysis/expectations/model-multiple-seasonal-cycles","title":"Account for daily and weekly patterns together","criterion":"Expectations use only earlier observations and expose the contribution and available history of each modeled cycle.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/expectations/model-multiple-seasonal-cycles, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","order":10}},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","status":"planned","area":"analysis","path":".capabilities/analysis/rates/assess-bounded-count-rates.capability.yaml","health":"planned","coverage":"uncovered","summary":"Analyze explicitly declared success-count and eligible-count rates with denominator-aware evidence and a stated count-model assumption.","intent":"Distinguish a sparse conversion swing from a well-supported change without losing the numerator and denominator behind the rate.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"analysis/rates/assess-bounded-count-rates has no automated or manual verification checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-automated-checks","message":"analysis/rates/assess-bounded-count-rates has no automated checks","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"},{"code":"missing-manual-review","message":"analysis/rates/assess-bounded-count-rates has no manual review guidance","capabilityId":"analysis/rates/assess-bounded-count-rates","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\analysis\\rates\\assess-bounded-count-rates.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Callers declare aligned success and eligible counts; non-integer, negative, or success-above-eligible inputs are rejected for the bounded-count model.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Evidence retains numerator, denominator, rate, and the declared reference population rather than reporting a percentage alone.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."},{"capabilityId":"analysis/rates/assess-bounded-count-rates","title":"Interpret rate changes with their supporting volume","criterion":"Statistical rate departures identify the model and support used and remain distinct from exact caller-supplied rate-limit violations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to analysis/rates/assess-bounded-count-rates, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Judge rates using their exposure","order":20}},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","status":"planned","area":"investigation","path":".capabilities/investigation/regimes/propose-reviewed-baseline-changes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Propose baseline-change candidates from sustained evidence and compare existing and proposed regimes before a caller explicitly accepts a reset.","intent":"Help distinguish an unresolved departure from a reviewed process change without automatically normalizing away an incident.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated or manual verification checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/regimes/propose-reviewed-baseline-changes has no automated checks","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/regimes/propose-reviewed-baseline-changes has no manual review guidance","capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\regimes\\propose-reviewed-baseline-changes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A preview shows old and proposed baselines, calibration support, and findings affected by the reset, including insufficient history after the boundary.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"A proposal records the candidate boundary, first detection time, supporting observations, and limitations; estimated onset is distinct from detection time.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/regimes/propose-reviewed-baseline-changes","title":"Review a possible new baseline before accepting it","criterion":"Only an explicit caller acceptance produces a new request with a declared reset; rejected proposals remain inspectable and have no numerical effect.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/regimes/propose-reviewed-baseline-changes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Business-aware expectations","backbone":"Make expectations fit the process","step":"Review a proposed regime transition","order":30}}],"deliveryStrategy":{"release":"Future - Business-aware expectations","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["analysis/expectations/model-multiple-seasonal-cycles"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"analysis/expectations/model-multiple-seasonal-cycles has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Make expectations fit the process"],"missingBackbones":[],"stepCoverageGaps":["Make expectations fit the process > Judge rates using their exposure","Make expectations fit the process > Review a proposed regime transition"],"rationale":"Starts with the earliest mapped step in Make expectations fit the process."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["analysis/rates/assess-bounded-count-rates","investigation/regimes/propose-reviewed-baseline-changes"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"analysis/rates/assess-bounded-count-rates has verification gaps; investigation/regimes/propose-reviewed-baseline-changes has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Make expectations fit the process"],"missingBackbones":[],"stepCoverageGaps":["Make expectations fit the process > Represent more than one repeating cycle"],"rationale":"Follows story-map order across Judge rates using their exposure, Review a proposed regime transition."}]},"presentation":{"outcome":"Future - Business-aware expectations helps teams move from Represent more than one repeating cycle to Review a proposed regime transition across 1 backbone activity.","narrativePath":[{"backbone":"Make expectations fit the process","step":"Represent more than one repeating cycle","capabilityIds":["analysis/expectations/model-multiple-seasonal-cycles"],"health":"planned"},{"backbone":"Make expectations fit the process","step":"Judge rates using their exposure","capabilityIds":["analysis/rates/assess-bounded-count-rates"],"health":"planned"},{"backbone":"Make expectations fit the process","step":"Review a proposed regime transition","capabilityIds":["investigation/regimes/propose-reviewed-baseline-changes"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Calendar and data readiness","capabilities":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","status":"planned","area":"data","path":".capabilities/data/calendar/aggregate-local-calendar-periods.capability.yaml","health":"planned","coverage":"uncovered","summary":"Aggregate completed local calendar days or weeks using an explicit timezone and report the actual boundaries and supporting observations.","intent":"Align business-period comparisons with the calendar people use, including daylight-saving transitions.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/calendar/aggregate-local-calendar-periods has no automated or manual verification checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-automated-checks","message":"data/calendar/aggregate-local-calendar-periods has no automated checks","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"},{"code":"missing-manual-review","message":"data/calendar/aggregate-local-calendar-periods has no manual review guidance","capabilityId":"data/calendar/aggregate-local-calendar-periods","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\calendar\\aggregate-local-calendar-periods.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Callers select local-day or local-week aggregation, an IANA timezone, a week-start rule when needed, and sum, mean, or last semantics.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Completeness is determined from the declared native cadence and source completeness; missing slots and unfinished periods remain explicit.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."},{"capabilityId":"data/calendar/aggregate-local-calendar-periods","title":"Compare days and weeks in my business timezone","criterion":"Existing fixed-duration analysis remains reproducible, and ambiguous local input timestamps still require caller disambiguation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/calendar/aggregate-local-calendar-periods, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","order":10}},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","status":"planned","area":"data","path":".capabilities/data/quality/report-source-freshness.capability.yaml","health":"planned","coverage":"uncovered","summary":"Evaluate caller-declared update deadlines and allowed lateness separately from value anomalies and propagate stale-source limitations to dependent evidence.","intent":"Prevent an old or absent observation from being interpreted as current healthy behavior or a confirmed business change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/quality/report-source-freshness has no automated or manual verification checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-automated-checks","message":"data/quality/report-source-freshness has no automated checks","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"},{"code":"missing-manual-review","message":"data/quality/report-source-freshness has no manual review guidance","capabilityId":"data/quality/report-source-freshness","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\quality\\report-source-freshness.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"A stale source leaves its historical numerical evidence available while marking dependent current relationship conclusions as stale or unavailable.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Callers supply an evaluation time, expected update schedule, timezone when relevant, and allowed lateness for each source.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."},{"capabilityId":"data/quality/report-source-freshness","title":"Know when missing updates weaken the conclusion","criterion":"Re-evaluating freshness at another supplied time preserves historical numerical evidence and records the changed freshness decision.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/quality/report-source-freshness, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Calendar and data readiness","backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","order":20}}],"deliveryStrategy":{"release":"Future - Calendar and data readiness","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/calendar/aggregate-local-calendar-periods"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/calendar/aggregate-local-calendar-periods has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Monitor the periods my business uses"],"missingBackbones":[],"stepCoverageGaps":["Monitor the periods my business uses > Check whether the sources are current"],"rationale":"Starts with the earliest mapped step in Monitor the periods my business uses."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["data/quality/report-source-freshness"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"data/quality/report-source-freshness has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Monitor the periods my business uses"],"missingBackbones":[],"stepCoverageGaps":["Monitor the periods my business uses > Declare local calendar boundaries"],"rationale":"Follows story-map order across Check whether the sources are current."}]},"presentation":{"outcome":"Future - Calendar and data readiness helps teams move from Declare local calendar boundaries to Check whether the sources are current across 1 backbone activity.","narrativePath":[{"backbone":"Monitor the periods my business uses","step":"Declare local calendar boundaries","capabilityIds":["data/calendar/aggregate-local-calendar-periods"],"health":"planned"},{"backbone":"Monitor the periods my business uses","step":"Check whether the sources are current","capabilityIds":["data/quality/report-source-freshness"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Cohort investigation","capabilities":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","status":"planned","area":"data","path":".capabilities/data/cohorts/analyze-declared-entity-batches.capability.yaml","health":"planned","coverage":"uncovered","summary":"Run a bounded batch of explicitly partitioned entity series and expose per-entity findings, support, and failures in one reviewable result.","intent":"Help callers inspect customers, regions, or services without collapsing their distinct baselines into one misleading series.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/cohorts/analyze-declared-entity-batches has no automated or manual verification checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-automated-checks","message":"data/cohorts/analyze-declared-entity-batches has no automated checks","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"},{"code":"missing-manual-review","message":"data/cohorts/analyze-declared-entity-batches has no manual review guidance","capabilityId":"data/cohorts/analyze-declared-entity-batches","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\cohorts\\analyze-declared-entity-batches.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Callers declare entity keys, metric units, cadence, and a batch resource budget; each entity receives a separate validated analysis request.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"Results report the evaluated population and raw finding counts without implying that many simultaneous tests establish a confirmed incident.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."},{"capabilityId":"data/cohorts/analyze-declared-entity-batches","title":"Analyze many entities without mixing their histories","criterion":"The result preserves each entity's findings, insufficient-history state, data quality, and failure status; one failure leaves healthy entities available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/cohorts/analyze-declared-entity-batches, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","order":10}},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","status":"planned","area":"investigation","path":".capabilities/investigation/cohorts/explain-additive-segment-contributions.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose an aggregate departure into contributions from caller-declared disjoint additive segments with compatible periods and units.","intent":"Direct investigation toward the parts of a total that changed while keeping arithmetic contribution distinct from causal explanation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/cohorts/explain-additive-segment-contributions has no automated or manual verification checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/cohorts/explain-additive-segment-contributions has no automated checks","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"},{"code":"missing-manual-review","message":"investigation/cohorts/explain-additive-segment-contributions has no manual review guidance","capabilityId":"investigation/cohorts/explain-additive-segment-contributions","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\cohorts\\explain-additive-segment-contributions.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Callers declare an additive total and disjoint segments with the same unit, boundaries, and entity scope; incompatible or overlapping segments are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Missing or uncovered segments appear as an explicit remainder or unavailable decomposition; contribution shares are unavailable for a zero total change.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."},{"capabilityId":"investigation/cohorts/explain-additive-segment-contributions","title":"Find which segments contribute to an aggregate change","criterion":"Reports show observed segment changes against the declared comparison reference and reconcile their sum to the total change within stated tolerance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/cohorts/explain-additive-segment-contributions, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Locate contributors to a total change","order":20}},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","status":"planned","area":"relationships","path":".capabilities/relationships/cohorts/separate-rate-and-population-mix.capability.yaml","health":"planned","coverage":"uncovered","summary":"Decompose a declared aggregate rate change into within-cohort rate and cohort-weight changes using explicit aligned counts and a comparison period.","intent":"Explain aggregate conversion movements that can occur even when individual cohorts' performance is stable or improving.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"relationships/cohorts/separate-rate-and-population-mix has no automated or manual verification checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-automated-checks","message":"relationships/cohorts/separate-rate-and-population-mix has no automated checks","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"},{"code":"missing-manual-review","message":"relationships/cohorts/separate-rate-and-population-mix has no manual review guidance","capabilityId":"relationships/cohorts/separate-rate-and-population-mix","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\relationships\\cohorts\\separate-rate-and-population-mix.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Callers supply disjoint cohort numerators and denominators for explicit reference and comparison periods; invalid counts or overlap are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Empty, missing, new, or disappearing cohorts retain explicit support limits; unavailable components are not filled with zero or presented as causes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."},{"capabilityId":"relationships/cohorts/separate-rate-and-population-mix","title":"Distinguish changing performance from a changing population mix","criterion":"Rate and mix components reconcile to the aggregate change within stated tolerance; interaction allocation is visible rather than hidden.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to relationships/cohorts/separate-rate-and-population-mix, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Cohort investigation","backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","order":30}}],"deliveryStrategy":{"release":"Future - Cohort investigation","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/cohorts/analyze-declared-entity-batches"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/cohorts/analyze-declared-entity-batches has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Investigate changes across populations"],"missingBackbones":[],"stepCoverageGaps":["Investigate changes across populations > Locate contributors to a total change","Investigate changes across populations > Separate performance changes from population mix"],"rationale":"Starts with the earliest mapped step in Investigate changes across populations."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["investigation/cohorts/explain-additive-segment-contributions","relationships/cohorts/separate-rate-and-population-mix"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"investigation/cohorts/explain-additive-segment-contributions has verification gaps; relationships/cohorts/separate-rate-and-population-mix has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Investigate changes across populations"],"missingBackbones":[],"stepCoverageGaps":["Investigate changes across populations > Keep each entity's evidence separate"],"rationale":"Follows story-map order across Locate contributors to a total change, Separate performance changes from population mix."}]},"presentation":{"outcome":"Future - Cohort investigation helps teams move from Keep each entity's evidence separate to Separate performance changes from population mix across 1 backbone activity.","narrativePath":[{"backbone":"Investigate changes across populations","step":"Keep each entity's evidence separate","capabilityIds":["data/cohorts/analyze-declared-entity-batches"],"health":"planned"},{"backbone":"Investigate changes across populations","step":"Locate contributors to a total change","capabilityIds":["investigation/cohorts/explain-additive-segment-contributions"],"health":"planned"},{"backbone":"Investigate changes across populations","step":"Separate performance changes from population mix","capabilityIds":["relationships/cohorts/separate-rate-and-population-mix"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Continuous analysis","capabilities":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","status":"planned","area":"data","path":".capabilities/data/revisions/reconcile-late-and-corrected-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Accept explicit source revisions, recompute affected analysis and relationships, and retain the differences from previously published evidence.","intent":"Let callers correct history without silently overwriting what earlier analysis actually showed.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"data/revisions/reconcile-late-and-corrected-observations has no automated or manual verification checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-automated-checks","message":"data/revisions/reconcile-late-and-corrected-observations has no automated checks","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"},{"code":"missing-manual-review","message":"data/revisions/reconcile-late-and-corrected-observations has no manual review guidance","capabilityId":"data/revisions/reconcile-late-and-corrected-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\data\\revisions\\reconcile-late-and-corrected-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"A revision identifies the source, affected timestamps or positions, prior source version, replacement data, and declared completeness changes.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Invalid or conflicting revisions leave the previous state available; unchanged inputs preserve numerical evidence and unrelated source results.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."},{"capabilityId":"data/revisions/reconcile-late-and-corrected-observations","title":"See how late or corrected data changes the evidence","criterion":"Reanalysis includes downstream reference, calibration, pattern, and relationship effects rather than changing only the corrected sample.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to data/revisions/reconcile-late-and-corrected-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","order":30}},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","status":"planned","area":"execution","path":".capabilities/execution/incremental/append-finalized-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Append finalized observations to supported analysis state and return new evidence without reanalyzing unchanged history.","intent":"Reduce repeated work for caller-managed monitoring while preserving causal scoring and continuity of sustained-pattern evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/incremental/append-finalized-observations has no automated or manual verification checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-automated-checks","message":"execution/incremental/append-finalized-observations has no automated checks","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"},{"code":"missing-manual-review","message":"execution/incremental/append-finalized-observations has no manual review guidance","capabilityId":"execution/incremental/append-finalized-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\incremental\\append-finalized-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Callers append observations after the stored cutoff using unchanged source identity and settings; backdated changes require an explicit revision workflow.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Outputs identify newly available evidence, current cutoff, remaining support limitations, and resource-limit stops without silently skipping observations.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."},{"capabilityId":"execution/incremental/append-finalized-observations","title":"Update evidence when the next observations arrive","criterion":"Retrying an identified append batch does not duplicate samples or findings; failed appends leave the prior committed state available.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/incremental/append-finalized-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","order":20}},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","status":"planned","area":"execution","path":".capabilities/execution/state/save-and-restore-analysis-state.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export and restore versioned local analysis state with enough provenance to continue supported recipes without losing calibration or pattern history.","intent":"Let calling applications maintain continuity across process restarts while preserving reproducible evidence.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"execution/state/save-and-restore-analysis-state has no automated or manual verification checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-automated-checks","message":"execution/state/save-and-restore-analysis-state has no automated checks","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"},{"code":"missing-manual-review","message":"execution/state/save-and-restore-analysis-state has no manual review guidance","capabilityId":"execution/state/save-and-restore-analysis-state","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\execution\\state\\save-and-restore-analysis-state.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"A snapshot records its state format, engine version, supported recipe, settings, source identity, last finalized position, and numerical provenance.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Export requires an explicit caller destination and protects existing files; state retention and deletion remain caller-controlled.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."},{"capabilityId":"execution/state/save-and-restore-analysis-state","title":"Resume local analysis after a restart","criterion":"Restore checks integrity and compatibility and rejects corrupt, unsupported, or mismatched state with a structured reason.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to execution/state/save-and-restore-analysis-state, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Continuous analysis","backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","order":10}}],"deliveryStrategy":{"release":"Future - Continuous analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["execution/state/save-and-restore-analysis-state"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"execution/state/save-and-restore-analysis-state has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Keep evidence current as observations arrive"],"missingBackbones":[],"stepCoverageGaps":["Keep evidence current as observations arrive > Analyze the next finalized samples","Keep evidence current as observations arrive > Reconcile revisions without losing history"],"rationale":"Starts with the earliest mapped step in Keep evidence current as observations arrive."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["execution/incremental/append-finalized-observations","data/revisions/reconcile-late-and-corrected-observations"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"data/revisions/reconcile-late-and-corrected-observations has verification gaps; execution/incremental/append-finalized-observations has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Keep evidence current as observations arrive"],"missingBackbones":[],"stepCoverageGaps":["Keep evidence current as observations arrive > Preserve state across restarts"],"rationale":"Follows story-map order across Analyze the next finalized samples, Reconcile revisions without losing history."}]},"presentation":{"outcome":"Future - Continuous analysis helps teams move from Preserve state across restarts to Reconcile revisions without losing history across 1 backbone activity.","narrativePath":[{"backbone":"Keep evidence current as observations arrive","step":"Preserve state across restarts","capabilityIds":["execution/state/save-and-restore-analysis-state"],"health":"planned"},{"backbone":"Keep evidence current as observations arrive","step":"Analyze the next finalized samples","capabilityIds":["execution/incremental/append-finalized-observations"],"health":"planned"},{"backbone":"Keep evidence current as observations arrive","step":"Reconcile revisions without losing history","capabilityIds":["data/revisions/reconcile-late-and-corrected-observations"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Historical evaluation","capabilities":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","status":"planned","area":"evaluation","path":".capabilities/evaluation/comparisons/compare-monitor-configurations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Compare a bounded caller-supplied set of recipes, thresholds, and policies on the same historical replay and a separate evaluation interval.","intent":"Make the tradeoff between earlier detection and unnecessary investigation inspectable before a caller adopts new settings.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/comparisons/compare-monitor-configurations has no automated or manual verification checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/comparisons/compare-monitor-configurations has no automated checks","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/comparisons/compare-monitor-configurations has no manual review guidance","capabilityId":"evaluation/comparisons/compare-monitor-configurations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\comparisons\\compare-monitor-configurations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Callers declare candidate configurations, a selection interval, and a later non-overlapping evaluation interval; overlapping intervals are rejected.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Candidate reports use the same cutoffs, data, label coverage, and matching rules and expose failed or inapplicable candidates.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/comparisons/compare-monitor-configurations","title":"Choose settings using held-out historical evidence","criterion":"Comparisons report finding volume, reviewed-incident coverage, detection delay, and policy eligibility without merging overlapping evidence as independent wins.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/comparisons/compare-monitor-configurations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","order":30}},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","status":"planned","area":"evaluation","path":".capabilities/evaluation/outcomes/compare-findings-with-reviewed-incidents.capability.yaml","health":"planned","coverage":"uncovered","summary":"Match historical findings to caller-reviewed incident intervals and report detection timing and unmatched findings under explicit matching rules.","intent":"Help owners judge usefulness on representative history rather than treating synthetic examples or trigger counts as detection accuracy.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated or manual verification checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no automated checks","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/outcomes/compare-findings-with-reviewed-incidents has no manual review guidance","capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\outcomes\\compare-findings-with-reviewed-incidents.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Precision or recall is reported only for an explicitly reviewed evaluation population, with its counts and matching rules; unlabeled periods stay unknown.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reports distinguish matched incidents, missed labeled incidents, duplicate findings, unmatched findings, and periods without reviewed labels.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/outcomes/compare-findings-with-reviewed-incidents","title":"Measure detection timing against reviewed incidents","criterion":"Reviewed labels declare entity, incident interval, and matching tolerance; invalid intervals and ambiguous matches are surfaced for review.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/outcomes/compare-findings-with-reviewed-incidents, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","order":20}},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","status":"planned","area":"evaluation","path":".capabilities/evaluation/replay/replay-historical-observations.capability.yaml","health":"planned","coverage":"uncovered","summary":"Replay a supplied history through explicit observation cutoffs and retain the evidence available at each cutoff.","intent":"Establish when a change could first have been seen without using later observations to improve earlier results.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"evaluation/replay/replay-historical-observations has no automated or manual verification checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-automated-checks","message":"evaluation/replay/replay-historical-observations has no automated checks","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"},{"code":"missing-manual-review","message":"evaluation/replay/replay-historical-observations has no manual review guidance","capabilityId":"evaluation/replay/replay-historical-observations","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\evaluation\\replay\\replay-historical-observations.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Callers supply a history, fixed analysis settings, and ordered replay cutoffs; every replay result records its cutoff and configuration.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Each cutoff excludes later observations from season selection, training, calibration, relationship fitting, and evaluation.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."},{"capabilityId":"evaluation/replay/replay-historical-observations","title":"Replay history as it would have arrived","criterion":"Incomplete, missing, and insufficient-history states remain visible at each cutoff rather than being replaced by the final run's evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to evaluation/replay/replay-historical-observations, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Historical evaluation","backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","order":10}}],"deliveryStrategy":{"release":"Future - Historical evaluation","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["evaluation/replay/replay-historical-observations"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"evaluation/replay/replay-historical-observations has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Evaluate a monitor before relying on it"],"missingBackbones":[],"stepCoverageGaps":["Evaluate a monitor before relying on it > Compare detection with reviewed outcomes","Evaluate a monitor before relying on it > Review the tradeoff before adopting settings"],"rationale":"Starts with the earliest mapped step in Evaluate a monitor before relying on it."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["evaluation/outcomes/compare-findings-with-reviewed-incidents","evaluation/comparisons/compare-monitor-configurations"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"evaluation/comparisons/compare-monitor-configurations has verification gaps; evaluation/outcomes/compare-findings-with-reviewed-incidents has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Evaluate a monitor before relying on it"],"missingBackbones":[],"stepCoverageGaps":["Evaluate a monitor before relying on it > Reconstruct the evidence available at the time"],"rationale":"Follows story-map order across Compare detection with reviewed outcomes, Review the tradeoff before adopting settings."}]},"presentation":{"outcome":"Future - Historical evaluation helps teams move from Reconstruct the evidence available at the time to Review the tradeoff before adopting settings across 1 backbone activity.","narrativePath":[{"backbone":"Evaluate a monitor before relying on it","step":"Reconstruct the evidence available at the time","capabilityIds":["evaluation/replay/replay-historical-observations"],"health":"planned"},{"backbone":"Evaluate a monitor before relying on it","step":"Compare detection with reviewed outcomes","capabilityIds":["evaluation/outcomes/compare-findings-with-reviewed-incidents"],"health":"planned"},{"backbone":"Evaluate a monitor before relying on it","step":"Review the tradeoff before adopting settings","capabilityIds":["evaluation/comparisons/compare-monitor-configurations"],"health":"planned"}],"coverageSignals":[]}},{"release":"Future - Investigation memory","capabilities":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","status":"planned","area":"interfaces","path":".capabilities/interfaces/export/package-investigation-evidence.capability.yaml","health":"planned","coverage":"uncovered","summary":"Export a local portable investigation bundle containing selected evidence, analysis provenance, limitations, and supplied review records.","intent":"Make an investigation reviewable beyond a transient MCP session or an agent explanation without requiring a hosted service.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"interfaces/export/package-investigation-evidence has no automated or manual verification checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-automated-checks","message":"interfaces/export/package-investigation-evidence has no automated checks","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"},{"code":"missing-manual-review","message":"interfaces/export/package-investigation-evidence has no manual review guidance","capabilityId":"interfaces/export/package-investigation-evidence","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\export\\package-investigation-evidence.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"A readable local report links to included structured evidence and clearly distinguishes observations, statistical findings, and caller review judgments.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Bundles contain a manifest, settings and engine versions, source fingerprints, selected numerical evidence, readiness, execution status, and lineage.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."},{"capabilityId":"interfaces/export/package-investigation-evidence","title":"Share an investigation that can be inspected later","criterion":"Callers explicitly select runs or cases and an output destination; existing output requires explicit overwrite permission.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to interfaces/export/package-investigation-evidence, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","order":20}},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/record-reviewed-case-outcomes.capability.yaml","health":"planned","coverage":"uncovered","summary":"Attach caller-authored review decisions, context, and observed outcomes to versioned case evidence without rewriting numerical findings.","intent":"Build an inspectable record of which findings mattered and what reviewers learned for later replay and investigation.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/record-reviewed-case-outcomes has no automated or manual verification checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/record-reviewed-case-outcomes has no automated checks","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/record-reviewed-case-outcomes has no manual review guidance","capabilityId":"investigation/history/record-reviewed-case-outcomes","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\record-reviewed-case-outcomes.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"A review record identifies the case and evidence version, reviewer-supplied identity, review time, disposition, and supporting notes or references.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Corrections create an attributable review revision while retaining the earlier decision and original numerical evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/record-reviewed-case-outcomes","title":"Keep reviewed outcomes beside the original evidence","criterion":"Records distinguish confirmed incidents, expected changes, data issues, dismissed concerns, and unresolved cases without requiring a causal label.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/record-reviewed-case-outcomes, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","order":10}},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","status":"planned","area":"investigation","path":".capabilities/investigation/history/retrieve-comparable-reviewed-cases.capability.yaml","health":"planned","coverage":"uncovered","summary":"Search caller-selected local case records using explicit entity, metric, pattern, and review filters and show why returned cases are comparable.","intent":"Give investigators useful precedent while keeping similarity separate from a diagnosis of the current change.","references":{"total":0,"readable":0,"missing":[]},"verification":{"automated":0,"manual":0,"gaps":[{"code":"missing-verification","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated or manual verification checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-automated-checks","message":"investigation/history/retrieve-comparable-reviewed-cases has no automated checks","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"},{"code":"missing-manual-review","message":"investigation/history/retrieve-comparable-reviewed-cases has no manual review guidance","capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\investigation\\history\\retrieve-comparable-reviewed-cases.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":4,"assessor-limitation":0,"ignored":0},"nextAction":"Keep as roadmap context until implementation starts.","topFindings":[{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Missing context, incompatible units, and insufficient baseline support are visible and cannot be silently ranked as equivalent evidence.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Results expose the matched attributes, material differences, evidence versions, and review dispositions behind the comparison.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."},{"capabilityId":"investigation/history/retrieve-comparable-reviewed-cases","title":"Find earlier reviewed cases with comparable evidence","criterion":"Retrieval leaves the current result and policy unchanged and returns an explicit no-comparable-case result when the supplied population has no match.","evidence":[],"status":"no-implementation-reference","confidence":"high","action":"add-implementation-reference","rationale":"The capability has no agent.implementation.references, so the assessor has no concrete files to inspect.","recommendation":"Add agent.implementation.references to investigation/history/retrieve-comparable-reviewed-cases, or lower the status if this capability is only planned."}],"storyMap":{"release":"Future - Investigation memory","backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","order":30}}],"deliveryStrategy":{"release":"Future - Investigation memory","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["investigation/history/record-reviewed-case-outcomes"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"investigation/history/record-reviewed-case-outcomes has verification gaps","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Reuse what an investigation teaches us"],"missingBackbones":[],"stepCoverageGaps":["Reuse what an investigation teaches us > Inspect comparable reviewed history","Reuse what an investigation teaches us > Preserve a portable evidence record"],"rationale":"Starts with the earliest mapped step in Reuse what an investigation teaches us."},{"order":2,"phase":"mid-game","name":"Progressive capability layer","capabilityIds":["interfaces/export/package-investigation-evidence","investigation/history/retrieve-comparable-reviewed-cases"],"releaseStrategy":"Add the next coherent layer of user-visible capability after the walking skeleton is usable.","developmentStrategy":"Build in story-map order while preserving the release narrative and avoiding isolated deep work.","riskIntent":"interfaces/export/package-investigation-evidence has verification gaps; investigation/history/retrieve-comparable-reviewed-cases has verification gaps","learningIntent":"Learn which adjacent steps increase outcome value before investing in optimization or hardening.","backboneCoverage":["Reuse what an investigation teaches us"],"missingBackbones":[],"stepCoverageGaps":["Reuse what an investigation teaches us > Record the reviewed outcome"],"rationale":"Follows story-map order across Inspect comparable reviewed history, Preserve a portable evidence record."}]},"presentation":{"outcome":"Future - Investigation memory helps teams move from Record the reviewed outcome to Inspect comparable reviewed history across 1 backbone activity.","narrativePath":[{"backbone":"Reuse what an investigation teaches us","step":"Record the reviewed outcome","capabilityIds":["investigation/history/record-reviewed-case-outcomes"],"health":"planned"},{"backbone":"Reuse what an investigation teaches us","step":"Preserve a portable evidence record","capabilityIds":["interfaces/export/package-investigation-evidence"],"health":"planned"},{"backbone":"Reuse what an investigation teaches us","step":"Inspect comparable reviewed history","capabilityIds":["investigation/history/retrieve-comparable-reviewed-cases"],"health":"planned"}],"coverageSignals":[]}},{"release":"Incomplete-period monitoring","capabilities":[{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","status":"implemented","area":"analysis","path":".capabilities/analysis/resolution/monitor-within-unfinished-days.capability.yaml","health":"review","coverage":"partial","summary":"Multi-resolution analysis keeps native finalized subperiod evidence and completed fixed 24-hour aggregates, with a separate marker for the unfinished day.","intent":"Expose changes at the supplied sample cadence before a chart's daily total is complete.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":5,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"A source cadence that does not evenly divide 24 hours is rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"A trailing partial aggregate reports observed progress and links its native evidence while remaining outside the completed-period series.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/resolution/monitor-within-unfinished-days","title":"See changes before the day is complete","criterion":"An explicitly unfinished source sample is excluded from both analysis views.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/resolution/monitor-within-unfinished-days --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","order":20}},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","status":"implemented","area":"data","path":".capabilities/data/quality/analyze-sparse-observations.capability.yaml","health":"review","coverage":"partial","summary":"Schema-1.1 callers declare coverage and missing-data semantics per source, keeping unknown collection distinct from confirmed empty additive volume.","intent":"Retain useful rule and supported statistical evidence for sparse histories while preserving time, provenance, incomplete periods and evidence limits.","references":{"total":10,"readable":10,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Run semantic review, save review evidence, or ignore accepted findings.","topFindings":[{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"As-of replay respects supplied arrival/revision timestamps and sparse multi-resolution views retain useful siblings without inventing complete totals.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":13,"excerpt":"\"seasonal-residual-v1\", \"adaptive-seasonal-v1\", \"multi-resolution-v1\","},{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":297,"excerpt":"if (self.data_policy or self.coverage or self.available_at or self.revisions) and self.timestamps is None:"},{"reference":"src/anomalyzer/contracts.py","line":298,"excerpt":"raise ValueError(\"source policies/coverage/revisions require original timestamps\")"},{"reference":"src/anomalyzer/contracts.py","line":437,"excerpt":"total_points = sum(len(dataset.values) + len(dataset.revisions) for dataset in self.datasets)"},{"reference":"src/anomalyzer/sparse.py","line":4,"excerpt":"Event coordinates retain actual timestamps and represent comparable events, not time."},{"reference":"src/anomalyzer/sparse.py","line":34,"excerpt":"if ds.data_policy or ds.coverage is not None or ds.available_at is not None or ds.revisions or any(v is None for v in ds.values) or \"incomplete\" in (ds.period_statuses or []):"},{"reference":"src/anomalyzer/sparse.py","line":64,"excerpt":"raise ValueError(\"duplicate timestamps require declared duplicate_policy; event_index needs unique event timestamps\")"},{"reference":"src/anomalyzer/sparse.py","line":66,"excerpt":"if coverage != \"complete\" or previous[\"coverage\"] != \"complete\" or ds.available_at or ds.revisions:"},{"reference":"src/anomalyzer/sparse.py","line":83,"excerpt":"observed_at = timestamp(revision.timestamp, ds.timezone)"},{"reference":"src/anomalyzer/sparse_views.py","line":1,"excerpt":"\"\"\"Schema-1.1 native and complete-day evidence with independent view eligibility.\"\"\""},{"reference":"src/anomalyzer/sparse_views.py","line":28,"excerpt":"case.revision_id = f\"{resolution}:{case.revision_id}\""},{"reference":"src/anomalyzer/sparse_views.py","line":29,"excerpt":"case.non_claims.append(\"Native and completed-day views reuse the same observations; they are not independent confirmation.\")"},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/orchestrator.py","line":106,"excerpt":"if request.config.recipe == \"multi-resolution-v1\":"},{"reference":"src/anomalyzer/orchestrator.py","line":107,"excerpt":"from .sparse_views import analyze_multi_request"},{"reference":"src/anomalyzer/orchestrator.py","line":133,"excerpt":"if \"data_policy\" in source.quality and request.config.recipe != \"multi-resolution-v1\""},{"reference":"src/anomalyzer/orchestrator.py","line":166,"excerpt":"if any(definitions[name].timestamps is None for name in _source_ids(relationship)):"},{"reference":"src/anomalyzer/orchestrator.py","line":180,"excerpt":"reset_positions=active_resets(request.config.reset_points, derived.timestamps or [], derived.timezone, request.context.as_of),"},{"reference":"src/anomalyzer/relationships.py","line":160,"excerpt":"source_timestamps = {source.dataset.id: source.times[source_index]"},{"reference":"src/anomalyzer/relationships.py","line":162,"excerpt":"cutoff = max(source_timestamps.values())"},{"reference":"src/anomalyzer/relationships.py","line":166,"excerpt":"source_indexes=source_indexes, source_timestamps=source_timestamps,"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":76,"excerpt":"(lambda d: d[\"timestamps\"].__setitem__(2, \"2026-01-01T02:30:00Z\"), \"off\"),"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":124,"excerpt":"def test_sparse_phase_without_support_is_explicitly_unavailable():"},{"reference":"tests/test_sparse.py","line":181,"excerpt":"ds[\"timestamps\"][1] = \"2026-02-01T00:00:00Z\""},{"reference":"docs/sparse-data.md","line":30,"excerpt":"timestamps=[\"2026-01-01T00:00:00Z\", \"2026-01-01T01:00:00Z\","},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":73,"excerpt":"conversion reasons, counts, revision IDs and availability timestamps. Converted"},{"reference":"docs/sparse-data.md","line":87,"excerpt":"otherwise bounds come from the supplied as-of-available timestamp range."},{"reference":"docs/sparse-data.md","line":105,"excerpt":"Duplicate timestamps normally fail. Legacy schema-1.1 relationship preparation"},{"reference":"examples/sparse_examples.py","line":59,"excerpt":"daily = request(dataset([1.0]*60, data_policy={\"measurement\":\"additive_volume\"}), recipe=\"multi-resolution-v1\","},{"reference":"examples/sparse_examples.py","line":95,"excerpt":"daily = next(item for item in outputs[\"unfinished_daily\"].dataset_results if item.resolution == \"completed_period\")"},{"reference":"examples/render_readme_progression.py","line":686,"excerpt":"def figure_multi_resolution():"},{"reference":"examples/render_readme_progression.py","line":699,"excerpt":"\"config\": {\"recipe\": \"multi-resolution-v1\", \"aggregate_function\": \"sum\"},"},{"reference":"examples/render_readme_progression.py","line":710,"excerpt":"assert partial[\"included_in_completed_period_view\"] is False"},{"reference":"examples/render_readme_progression.py","line":767,"excerpt":"complete_count = result.data_quality[\"views\"][\"completed_period\"][\"observation_count\"]"},{"reference":"examples/render_readme_progression.py","line":884,"excerpt":"header(svg, \"14\", \"Choose what an empty period means\", \"Identical source timestamps and values; only the declared empty-data policy changes.\")"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"Bounded calendar preparation preserves phase, original values and coverage; only an explicit additive-volume empty policy authorizes zero conversion.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":51,"excerpt":"def valid_reset_points(cls, values):"},{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":71,"excerpt":"def numeric_values(cls, data):"},{"reference":"src/anomalyzer/contracts.py","line":85,"excerpt":"if self.period_statuses is not None and len(self.period_statuses) != len(self.values):"},{"reference":"src/anomalyzer/contracts.py","line":86,"excerpt":"raise ValueError(\"period_statuses and values must have equal length\")"},{"reference":"src/anomalyzer/sparse.py","line":3,"excerpt":"Calendar coordinates remain dense only inside an explicitly bounded input grid."},{"reference":"src/anomalyzer/sparse.py","line":34,"excerpt":"if ds.data_policy or ds.coverage is not None or ds.available_at is not None or ds.revisions or any(v is None for v in ds.values) or \"incomplete\" in (ds.period_statuses or []):"},{"reference":"src/anomalyzer/sparse.py","line":42,"excerpt":"\"\"\"Return the explicit grid, eligible model values and inspectable provenance.\"\"\""},{"reference":"src/anomalyzer/sparse.py","line":48,"excerpt":"statuses = ds.period_statuses or [\"complete\"] * len(ds.values)"},{"reference":"src/anomalyzer/sparse.py","line":54,"excerpt":"\"complete\" if value is not None else \"empty\" if policy.absence_means_zero else \"unknown\")"},{"reference":"src/anomalyzer/sparse_views.py","line":72,"excerpt":"measurements = quality[\"rule_values\"]"},{"reference":"src/anomalyzer/sparse_views.py","line":75,"excerpt":"if policy.observation_basis != \"calendar\":"},{"reference":"src/anomalyzer/sparse_views.py","line":82,"excerpt":"if policy.measurement != \"additive_volume\" or cfg.aggregate_function != \"sum\":"},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":98,"excerpt":"daily_times, daily_values, coverage, records, available_at = [], [], [], [], []"},{"reference":"src/anomalyzer/orchestrator.py","line":28,"excerpt":"if not any(row.get(\"available_at\") and (row.get(\"value\") is not None or row.get(\"coverage\") == \"excluded\") for row in samples):"},{"reference":"src/anomalyzer/orchestrator.py","line":63,"excerpt":"unavailable values. Dropping only that leading run lets the ordinary engine"},{"reference":"src/anomalyzer/orchestrator.py","line":127,"excerpt":"prepared_points += len(source.values) + len(dataset.revisions)"},{"reference":"src/anomalyzer/orchestrator.py","line":139,"excerpt":"source.quality.get(\"rule_values\", source.values)[:source.quality.get(\"grid_count\", source.quality[\"observation_count\"])], dataset.rules,"},{"reference":"src/anomalyzer/orchestrator.py","line":191,"excerpt":"[dict(value=p.value, available_at=p.latest_source_cutoff if any(p.source_available_at.values()) else None) for p in lineage])"},{"reference":"src/anomalyzer/relationships.py","line":36,"excerpt":"sources = [replace(source, values=source.quality.get(\"rule_values\", source.values)) for source in sources]"},{"reference":"src/anomalyzer/relationships.py","line":58,"excerpt":"x = predictor.values[index - relationship.lag]"},{"reference":"src/anomalyzer/relationships.py","line":98,"excerpt":"value, unavailable = None, \"zero_denominator\""},{"reference":"src/anomalyzer/relationships.py","line":131,"excerpt":"y, x = response.values[index], predictor.values[predictor_index]"},{"reference":"src/anomalyzer/relationships.py","line":141,"excerpt":"source_values = [source.values[source_index]"},{"reference":"src/anomalyzer/early.py","line":58,"excerpt":"for index, (observed_at, value) in enumerate(zip(times, values)):"},{"reference":"src/anomalyzer/early.py","line":137,"excerpt":"criterion_met and policy.allow_explicit_rules)"},{"reference":"tests/test_sparse.py","line":45,"excerpt":"def test_zero_is_only_authorized_for_complete_empty_additive_volume():"},{"reference":"tests/test_sparse.py","line":46,"excerpt":"request = body([2, None, None, 0, 9], data_policy={\"missing\": \"zero_for_empty\", \"measurement\": \"additive_volume\"},"},{"reference":"tests/test_sparse.py","line":51,"excerpt":"assert [r[\"original_value\"] for r in rows] == [2, None, None, 0, 9]"},{"reference":"tests/test_sparse.py","line":52,"excerpt":"assert rows[1][\"reason\"] == \"confirmed_empty_additive_volume\""},{"reference":"tests/test_sparse.py","line":58,"excerpt":"def test_global_collection_contract_authorizes_omitted_slots_but_not_explicit_unknown():"},{"reference":"docs/sparse-data.md","line":4,"excerpt":"Schema 1.1 therefore preserves missing values by default. An expectation can be"},{"reference":"docs/sparse-data.md","line":33,"excerpt":"data_policy=DataPolicy(),  # preserve missing; hour 2 is also unknown"},{"reference":"docs/sparse-data.md","line":41,"excerpt":"| Numeric value, including zero, `coverage=\"complete\"` | A measurement eligible for rules and model support. Explicit zero remains zero. |"},{"reference":"docs/sparse-data.md","line":43,"excerpt":"| `coverage=\"empty\"` | Caller confirms a finalized period with no reported volume. Preserve as missing unless an additive zero policy explicitly authorizes conversion. |"},{"reference":"docs/sparse-data.md","line":44,"excerpt":"| `coverage=\"unknown\"` | Collection/exposure is unknown; never filled even under a global empty-volume contract. |"},{"reference":"examples/sparse_examples.py","line":22,"excerpt":"preserve = request(dataset(values, coverage=[\"empty\" if v is None else \"complete\" for v in values],"},{"reference":"examples/sparse_examples.py","line":23,"excerpt":"data_policy={\"measurement\": \"additive_volume\"}))"},{"reference":"examples/sparse_examples.py","line":25,"excerpt":"zero[\"datasets\"][0][\"data_policy\"][\"missing\"] = \"zero_for_empty\""},{"reference":"examples/sparse_examples.py","line":26,"excerpt":"result.update(preserve=preserve, zero_empty=zero)"},{"reference":"examples/sparse_examples.py","line":28,"excerpt":"coverage=[\"complete\"]*2+[\"unknown\"]+[\"complete\"]*6+[\"unknown\"]+[\"complete\"]*3+[\"incomplete\"], data_policy={}))"},{"reference":"examples/render_readme_progression.py","line":135,"excerpt":"def chart(svg, values, overlays, x, y, width, height, *, y_bounds=None,"},{"reference":"examples/render_readme_progression.py","line":139,"excerpt":"all_values = [values] + [series for _, series, *_ in overlays]"},{"reference":"examples/render_readme_progression.py","line":140,"excerpt":"low, high = y_bounds or bounds(*all_values)"},{"reference":"examples/render_readme_progression.py","line":166,"excerpt":"if value is not None and (index == 0 or values[index-1] is None) and (index == count-1 or values[index+1] is None):"},{"reference":"examples/render_readme_progression.py","line":206,"excerpt":"values = [80 + 0.65 * index + [0, 0.25, -0.15, 0.10, -0.20][index % 5]"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"data/quality/analyze-sparse-observations","title":"Assess sparse observations without inventing consumption","criterion":"Event-index profiles preserve real event timestamps and reject calendar seasonality; unavailable relationship inputs and zero denominators stay unavailable.","evidence":[{"reference":"src/anomalyzer/contracts.py","line":54,"excerpt":"raise ValueError(\"reset_points must contain nonempty timestamps or integer sample positions\")"},{"reference":"src/anomalyzer/contracts.py","line":81,"excerpt":"elif self.frequency is None and getattr(getattr(self, \"data_policy\", None), \"observation_basis\", None) != \"event_index\":"},{"reference":"src/anomalyzer/contracts.py","line":242,"excerpt":"missing: Literal[\"preserve\", \"zero_for_empty\"] = \"preserve\""},{"reference":"src/anomalyzer/contracts.py","line":245,"excerpt":"observation_basis: Literal[\"calendar\", \"event_index\"] = \"calendar\""},{"reference":"src/anomalyzer/contracts.py","line":258,"excerpt":"if self.observation_basis == \"event_index\" and ("},{"reference":"src/anomalyzer/sparse.py","line":4,"excerpt":"Event coordinates retain actual timestamps and represent comparable events, not time."},{"reference":"src/anomalyzer/sparse.py","line":49,"excerpt":"for index, (text, value, status) in enumerate(zip(ds.timestamps, ds.values, statuses)):"},{"reference":"src/anomalyzer/sparse.py","line":57,"excerpt":"available = timestamp(ds.available_at[index], ds.timezone) if ds.available_at and ds.available_at[index] else None"},{"reference":"src/anomalyzer/sparse.py","line":63,"excerpt":"if cfg.duplicate_policy == \"error\" or policy.observation_basis == \"event_index\":"},{"reference":"src/anomalyzer/sparse.py","line":64,"excerpt":"raise ValueError(\"duplicate timestamps require declared duplicate_policy; event_index needs unique event timestamps\")"},{"reference":"src/anomalyzer/sparse_views.py","line":76,"excerpt":"unsupported = \"completed-day aggregation is unavailable on event_index\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/sparse_views.py","line":128,"excerpt":"policy_zero_children=sum(quality[\"samples\"][i][\"reason\"] == \"confirmed_empty_additive_volume\" for i in indexes),"},{"reference":"src/anomalyzer/sparse_views.py","line":129,"excerpt":"source_input_indexes=[raw for i in indexes for raw in quality[\"samples\"][i][\"source_indexes\"]],"},{"reference":"src/anomalyzer/sparse_views.py","line":134,"excerpt":"timestamps=daily_times, values=daily_values, frequency=\"1d\", coverage=coverage, available_at=available_at,"},{"reference":"src/anomalyzer/orchestrator.py","line":37,"excerpt":"assessment.baseline[\"known_input_availability_cutoff\"] = cutoffs[assessment.index]"},{"reference":"src/anomalyzer/orchestrator.py","line":38,"excerpt":"assessment.assumptions.append(\"Availability cutoff is conservative across known prior inputs; unknown arrival history cannot establish delivery-time detection performance.\")"},{"reference":"src/anomalyzer/orchestrator.py","line":166,"excerpt":"if any(definitions[name].timestamps is None for name in _source_ids(relationship)):"},{"reference":"src/anomalyzer/orchestrator.py","line":178,"excerpt":"relationship.id, derived.timestamps or [], derived.values,"},{"reference":"src/anomalyzer/orchestrator.py","line":180,"excerpt":"reset_positions=active_resets(request.config.reset_points, derived.timestamps or [], derived.timezone, request.context.as_of),"},{"reference":"src/anomalyzer/relationships.py","line":40,"excerpt":"raise ValueError(\"relationship timestamps cannot be positional\")"},{"reference":"src/anomalyzer/relationships.py","line":57,"excerpt":"for index in range(relationship.lag, coefficient_cutoff + 1):"},{"reference":"src/anomalyzer/relationships.py","line":58,"excerpt":"x = predictor.values[index - relationship.lag]"},{"reference":"src/anomalyzer/relationships.py","line":96,"excerpt":"if relationship.zero_denominator == \"error\":"},{"reference":"src/anomalyzer/relationships.py","line":97,"excerpt":"raise ValueError(f\"zero denominator at aligned index {index}\")"},{"reference":"src/anomalyzer/early.py","line":140,"excerpt":"unavailable_rules = [dict(id=rule.id, reason=\"no_comparable_prior\" if previous is None else \"relative_change_from_zero\")"},{"reference":"src/anomalyzer/early.py","line":147,"excerpt":"target_id=target_id, index=index, timestamp=observed_at,"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":68,"excerpt":"{\"observation_basis\": \"event_index\", \"grid_start\": \"2026-01-01T00:00:00Z\"},"},{"reference":"tests/test_sparse.py","line":76,"excerpt":"(lambda d: d[\"timestamps\"].__setitem__(2, \"2026-01-01T02:30:00Z\"), \"off\"),"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":129,"excerpt":"assert any(row[\"reason\"] == \"no_training_support_for_calendar_phase\" for row in result.methods[0].diagnostics[\"unavailable\"])"},{"reference":"docs/sparse-data.md","line":30,"excerpt":"timestamps=[\"2026-01-01T00:00:00Z\", \"2026-01-01T01:00:00Z\","},{"reference":"docs/sparse-data.md","line":105,"excerpt":"Duplicate timestamps normally fail. Legacy schema-1.1 relationship preparation"},{"reference":"docs/sparse-data.md","line":110,"excerpt":"rejected. Event-index timestamps must be unique; ties are refused rather than"},{"reference":"docs/sparse-data.md","line":141,"excerpt":"original calendar indexes and timestamps. Post-gap learning requires an explicit"},{"reference":"docs/sparse-data.md","line":155,"excerpt":"## Comparable event-index observations"},{"reference":"examples/sparse_examples.py","line":26,"excerpt":"result.update(preserve=preserve, zero_empty=zero)"},{"reference":"examples/sparse_examples.py","line":36,"excerpt":"events = dataset(costs, units=\"dollars_per_call\", data_policy={\"observation_basis\": \"event_index\"})"},{"reference":"examples/sparse_examples.py","line":38,"excerpt":"events[\"timestamps\"] = [(datetime(2026,1,1,tzinfo=timezone.utc)+timedelta(hours=h)).isoformat() for h in hours]"},{"reference":"examples/sparse_examples.py","line":57,"excerpt":"relation[\"relationships\"] = [{\"id\":\"unit-cost\",\"kind\":\"ratio\",\"numerator\":\"series\",\"denominator\":\"calls\",\"units\":\"dollars_per_call\"}]"},{"reference":"examples/sparse_examples.py","line":81,"excerpt":"preserve, zero = [outputs[n].dataset_results[0] for n in (\"preserve\", \"zero_empty\")]"},{"reference":"examples/render_readme_progression.py","line":654,"excerpt":"svg.text(430, 467, \"Sample index (zero-based)\", 15, MUTED, anchor=\"middle\")"},{"reference":"examples/render_readme_progression.py","line":697,"excerpt":"\"timestamps\": [(origin + timedelta(hours=index)).isoformat()"},{"reference":"examples/render_readme_progression.py","line":884,"excerpt":"header(svg, \"14\", \"Choose what an empty period means\", \"Identical source timestamps and values; only the declared empty-data policy changes.\")"},{"reference":"examples/render_readme_progression.py","line":887,"excerpt":"shared_high = max(v for v in observed(\"preserve\")+expectation(\"preserve\")+observed(\"zero_empty\")+expectation(\"zero_empty\") if v is not None)*1.12"},{"reference":"examples/render_readme_progression.py","line":897,"excerpt":"header(svg, \"15\", \"Keep zero, unknown and unfinished distinct\", \"Blue zeros are actual measurements; gaps stay gaps; the orange open-period value is excluded.\")"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Assess sparse source data","order":30}},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","status":"implemented","area":"data","path":".capabilities/data/quality/exclude-incomplete-periods.capability.yaml","health":"review","coverage":"partial","summary":"Explicitly incomplete observations are reported as data quality evidence and excluded from completed-series analysis.","intent":"Prevent an unfinished period from appearing to be an unexplained drop.","references":{"total":8,"readable":8,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"A relationship with an unavailable latest source reports that unavailability rather than normal behavior.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"Incomplete source samples are excluded from expectations, calibration, point scoring, and completed aggregates.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"data/quality/exclude-incomplete-periods","title":"Avoid mistaking unfinished periods for drops","criterion":"Schema-1.0 incomplete observations form a trailing suffix; schema-1.1 source policies preserve internal and trailing incomplete positions as unavailable.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify data/quality/exclude-incomplete-periods --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Incomplete-period monitoring","backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","order":10}}],"deliveryStrategy":{"release":"Incomplete-period monitoring","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["data/quality/exclude-incomplete-periods"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"data/quality/exclude-incomplete-periods needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Monitor while periods are still open"],"missingBackbones":[],"stepCoverageGaps":["Monitor while periods are still open > Assess sparse source data","Monitor while periods are still open > Compare finalized subperiods"],"rationale":"Starts with the earliest mapped step in Monitor while periods are still open."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["analysis/resolution/monitor-within-unfinished-days","data/quality/analyze-sparse-observations"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"analysis/resolution/monitor-within-unfinished-days needs review; data/quality/analyze-sparse-observations needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Monitor while periods are still open"],"missingBackbones":[],"stepCoverageGaps":["Monitor while periods are still open > Keep unfinished data visible"],"rationale":"Prioritizes capabilities with delivery risk signals: analysis/resolution/monitor-within-unfinished-days, data/quality/analyze-sparse-observations."}]},"presentation":{"outcome":"Incomplete-period monitoring helps teams move from Keep unfinished data visible to Assess sparse source data across 1 backbone activity.","narrativePath":[{"backbone":"Monitor while periods are still open","step":"Keep unfinished data visible","capabilityIds":["data/quality/exclude-incomplete-periods"],"health":"review"},{"backbone":"Monitor while periods are still open","step":"Compare finalized subperiods","capabilityIds":["analysis/resolution/monitor-within-unfinished-days"],"health":"review"},{"backbone":"Monitor while periods are still open","step":"Assess sparse source data","capabilityIds":["data/quality/analyze-sparse-observations"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Monitor while periods are still open > Assess sparse source data","message":"Monitor while periods are still open > Assess sparse source data has needs-review coverage."},{"kind":"weak","label":"Monitor while periods are still open > Compare finalized subperiods","message":"Monitor while periods are still open > Compare finalized subperiods has needs-review coverage."},{"kind":"weak","label":"Monitor while periods are still open > Keep unfinished data visible","message":"Monitor while periods are still open > Keep unfinished data visible has needs-review coverage."}]}},{"release":"Python integration","capabilities":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/analyze-through-public-api.capability.yaml","health":"review","coverage":"partial","summary":"The public Python API accepts series or canonical requests and returns the same numerical analysis used by the CLI and MCP compatibility interfaces.","intent":"Embed local analysis in an application without changing the meaning of the evidence.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze accepts bare ordered values with separate validated settings and dispatches canonical schema 1.0 or 1.1 requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_relationships preserves the declared multi-dataset relationship semantics.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/python/analyze-through-public-api","title":"Embed local analysis in my application","criterion":"analyze_series preserves supplied settings and event-time context for a single dataset.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/python/analyze-through-public-api --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Call the public API","order":10}},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","status":"implemented","area":"interfaces","path":".capabilities/interfaces/python/install-versioned-library.capability.yaml","health":"action","coverage":"partial","summary":"Applications can install an identified Anomalyzer distribution and run the public analysis API locally with bundled dependencies.","intent":"Replace copied source files with a reviewable dependency that can be pinned and rolled back independently of application deployments.","references":{"total":11,"readable":11,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[{"code":"declared-gap","message":"interfaces/python/install-versioned-library: First immutable GitHub Release requires owner setup and a reviewed version tag.","capabilityId":"interfaces/python/install-versioned-library","filePath":"C:\\Users\\troym\\Git\\Anomalyizer\\.capabilities\\interfaces\\python\\install-versioned-library.capability.yaml"}]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Resolve or explicitly ignore verification gaps.","topFindings":[{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"A Linux Python 3.12 build bundles compatible binary dependencies and executes the consumer example with networking disabled.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/publish_release.py","line":43,"excerpt":"\"Linux/Python 3.12 Lambda consumer verified offline. \""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":6,"excerpt":"bundle=\"$root/.tmp/lambda-consumer\""},{"reference":"scripts/verify_lambda.sh","line":18,"excerpt":"python -m pip download --only-binary=:all: --require-hashes \\"},{"reference":"scripts/verify_lambda.sh","line":20,"excerpt":"python -m pip install --no-index --only-binary=:all: --require-hashes \\"},{"reference":".github/workflows/ci.yml","line":29,"excerpt":"- run: uv run python examples/generate.py"},{"reference":".github/workflows/ci.yml","line":32,"excerpt":"- run: uv run python examples/run_demo.py"},{"reference":".github/workflows/ci.yml","line":33,"excerpt":"- run: uv run python examples/run_agentic_demo.py"},{"reference":".github/workflows/ci.yml","line":53,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/ci.yml","line":54,"excerpt":"run: uv export --locked --no-dev --no-emit-project --output-file dist/runtime-dependencies.txt"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":58,"excerpt":"uv run python examples/run_demo.py"},{"reference":"docs/releases/README.md","line":59,"excerpt":"uv run python examples/run_agentic_demo.py"},{"reference":"docs/releases/README.md","line":75,"excerpt":"versions, and tests a bundled Linux x86_64/Python 3.12 consumer in AWS's Lambda"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Clean consumers outside the source checkout can install either distribution and exercise contracts, recipes, assessments, relationships, readiness and cases.","evidence":[{"reference":"scripts/verify_distribution.py","line":1,"excerpt":"\"\"\"Install wheel AND sdist in clean environments outside the source tree (requires uv).\"\"\""},{"reference":"scripts/verify_distribution.py","line":42,"excerpt":"run([\"uv\", \"pip\", \"install\", \"--python\", python, \"--no-deps\", artifact], outside, env=env)"},{"reference":"examples/library/guard_smoke.py","line":32,"excerpt":"analyze_relationships, get_case, replay_policy,"},{"reference":"examples/library/guard_smoke.py","line":49,"excerpt":"assert legacy.methods[0].diagnostics[\"detection_readiness\"]"},{"reference":"examples/library/guard_smoke.py","line":65,"excerpt":"\"relationships\": [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"cost\","},{"reference":"examples/library/guard_smoke.py","line":69,"excerpt":"result = analyze_relationships(request)"},{"reference":"examples/library/guard_smoke.py","line":76,"excerpt":"short = analyze_relationships({**request.model_dump(mode=\"json\"),"},{"reference":".github/workflows/ci.yml","line":59,"excerpt":"- name: Test built wheel and sdist outside checkout"},{"reference":"docs/releases/README.md","line":13,"excerpt":"| Source distribution (`.tar.gz`, or sdist) | Source needed to build the package. Useful for rebuilding and inspection; installation also needs the declared build backend. |"},{"reference":"docs/releases/README.md","line":208,"excerpt":"`analyze_series`, `analyze_relationships`, `get_case`, `replay_policy` (also"},{"reference":"docs/releases/README.md","line":211,"excerpt":"`anomalyzer.contracts`. See the self-contained"},{"reference":"docs/releases/README.md","line":213,"excerpt":"all APIs, recipes, point/pattern evidence, relationships, cases and policy"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"interfaces/python/install-versioned-library","title":"Install a versioned analysis library","criterion":"Version tags agree with package metadata and published release assets include SHA-256 checksums, source revision and pinned transitive dependencies.","evidence":[{"reference":"pyproject.toml","line":26,"excerpt":"dependencies = [\"pydantic>=2.10,<3\", \"typing-extensions>=4.12.2\", \"tzdata>=2024.1; sys_platform == 'win32'\"]"},{"reference":"pyproject.toml","line":33,"excerpt":"[project.optional-dependencies]"},{"reference":"scripts/release.py","line":34,"excerpt":"return hashlib.sha256(path.read_bytes()).hexdigest()"},{"reference":"scripts/release.py","line":81,"excerpt":"runtime = dist / \"runtime-dependencies.txt\""},{"reference":"scripts/release.py","line":82,"excerpt":"assert runtime.is_file(), \"export locked runtime dependencies first\""},{"reference":"scripts/release.py","line":85,"excerpt":"suffix = f\" --hash=sha256:{digest(wheel)}\\n\""},{"reference":"scripts/release.py","line":92,"excerpt":"assets = sorted(p for p in dist.iterdir() if p.is_file()"},{"reference":"scripts/publish_release.py","line":1,"excerpt":"\"\"\"Publish tested assets without replacing tags or existing assets (GitHub CLI auth).\"\"\""},{"reference":"scripts/publish_release.py","line":25,"excerpt":"remote_sha = json.loads(gh(\"api\", f\"repos/{REPOSITORY}/commits/{tag}\"))[\"sha\"]"},{"reference":"scripts/publish_release.py","line":26,"excerpt":"if remote_sha != revision:"},{"reference":"scripts/publish_release.py","line":27,"excerpt":"raise ValueError(\"Remote tag no longer points to the tested source revision\")"},{"reference":"scripts/publish_release.py","line":28,"excerpt":"endpoint = f\"repos/{REPOSITORY}/releases/tags/{tag}\""},{"reference":"scripts/verify_distribution.py","line":41,"excerpt":"dist / \"runtime-dependencies.txt\"], outside, env=env)"},{"reference":"scripts/verify_lambda.sh","line":25,"excerpt":"sha256sum *.whl > /output/wheelhouse-SHA256SUMS"},{"reference":"examples/library/guard_smoke.py","line":39,"excerpt":"assert anomalyzer.__version__ == importlib.metadata.version(\"anomalyzer\")"},{"reference":".github/workflows/ci.yml","line":44,"excerpt":"- name: Check release tag before building"},{"reference":".github/workflows/ci.yml","line":47,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/ci.yml","line":48,"excerpt":"run: python scripts/release.py check-version --tag \"$RELEASE_TAG\""},{"reference":".github/workflows/ci.yml","line":51,"excerpt":"- name: Check package metadata"},{"reference":".github/workflows/ci.yml","line":53,"excerpt":"- name: Export locked runtime dependencies"},{"reference":".github/workflows/release.yml","line":31,"excerpt":"name: release-dist-${{ github.sha }}"},{"reference":".github/workflows/release.yml","line":33,"excerpt":"- name: Stage tested assets for owner publication"},{"reference":".github/workflows/release.yml","line":36,"excerpt":"RELEASE_TAG: ${{ github.ref_name }}"},{"reference":".github/workflows/release.yml","line":37,"excerpt":"run: python scripts/publish_release.py --tag \"$RELEASE_TAG\" --revision \"$GITHUB_SHA\""},{"reference":"docs/releases/README.md","line":3,"excerpt":"Anomalyzer is a normal Python dependency. Version **0.2.0** is the first release"},{"reference":"docs/releases/README.md","line":12,"excerpt":"| Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |"},{"reference":"docs/releases/README.md","line":14,"excerpt":"| Version tag (`v0.2.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |"},{"reference":"docs/releases/README.md","line":15,"excerpt":"| Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |"},{"reference":"docs/releases/README.md","line":19,"excerpt":"it for package metadata and the public `__version__` imports it."},{"reference":"tests/test_release.py","line":1,"excerpt":"\"\"\"Release safety: reject wrong tags and refuse changes to published artifacts.\"\"\""},{"reference":"tests/test_release.py","line":21,"excerpt":"def test_release_refuses_mismatched_or_unsupported_tag(tag):"},{"reference":"tests/test_release.py","line":23,"excerpt":"release.check_tag(tag)"},{"reference":"tests/test_release.py","line":52,"excerpt":"def test_published_matching_release_is_noop(monkeypatch, tmp_path):"},{"reference":"tests/test_release.py","line":81,"excerpt":"\"version\": release.version(), \"source_revision\": \"a\" * 40, \"tag\": f\"v{release.version()}\","}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Python integration","backbone":"Embed analysis in my application","step":"Install a released dependency","order":20}}],"deliveryStrategy":{"release":"Python integration","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/python/analyze-through-public-api"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/python/analyze-through-public-api needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Embed analysis in my application"],"missingBackbones":[],"stepCoverageGaps":["Embed analysis in my application > Install a released dependency"],"rationale":"Starts with the earliest mapped step in Embed analysis in my application."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/python/install-versioned-library"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/python/install-versioned-library has verification gaps; interfaces/python/install-versioned-library needs action","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Embed analysis in my application"],"missingBackbones":[],"stepCoverageGaps":["Embed analysis in my application > Call the public API"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/python/install-versioned-library."}]},"presentation":{"outcome":"Python integration helps teams move from Call the public API to Install a released dependency across 1 backbone activity.","narrativePath":[{"backbone":"Embed analysis in my application","step":"Call the public API","capabilityIds":["interfaces/python/analyze-through-public-api"],"health":"review"},{"backbone":"Embed analysis in my application","step":"Install a released dependency","capabilityIds":["interfaces/python/install-versioned-library"],"health":"action"}],"coverageSignals":[{"kind":"weak","label":"Embed analysis in my application > Call the public API","message":"Embed analysis in my application > Call the public API has needs-review coverage."},{"kind":"weak","label":"Embed analysis in my application > Install a released dependency","message":"Embed analysis in my application > Install a released dependency has needs-action coverage."}]}},{"release":"Read-only MCP server","capabilities":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/connect-read-only-local-server.capability.yaml","health":"review","coverage":"partial","summary":"The local stdio MCP server advertises three read-only investigation tools with schemas and interpretation guidance.","intent":"Allow agent clients to discover and call the analysis tools using a clear local protocol contract.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Analysis tool calls return structured content together with compact text.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Invalid tool input and unknown session run IDs return structured errors rather than fabricated results.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/connect-read-only-local-server","title":"Give an agent read-only access to local analysis","criterion":"Legacy initialization and modern stateless discovery return supported protocol metadata and interpretation guidance.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/connect-read-only-local-server --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Connect a local server","order":10}},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","status":"implemented","area":"interfaces","path":".capabilities/interfaces/mcp/investigate-findings-and-evidence.capability.yaml","health":"review","coverage":"partial","summary":"Agents can analyze a request, page through its findings, and retrieve supporting samples, assessments, readiness, and relationship lineage.","intent":"Keep the initial agent response compact while making the supporting evidence inspectable.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"get_evidence exposes the selected finding's numerical support and relationship source lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"list_findings pages through cases, candidates, episodes, patterns, and data-quality findings.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"interfaces/mcp/investigate-findings-and-evidence","title":"Let an agent inspect the evidence behind a finding","criterion":"The analyze tool returns a compact headline, finding preview, and run ID for subsequent evidence requests.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify interfaces/mcp/investigate-findings-and-evidence --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Read-only MCP server","backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","order":20}}],"deliveryStrategy":{"release":"Read-only MCP server","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["interfaces/mcp/connect-read-only-local-server"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"interfaces/mcp/connect-read-only-local-server needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Give an agent access to local evidence"],"missingBackbones":[],"stepCoverageGaps":["Give an agent access to local evidence > Drill into findings and evidence"],"rationale":"Starts with the earliest mapped step in Give an agent access to local evidence."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["interfaces/mcp/investigate-findings-and-evidence"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"interfaces/mcp/investigate-findings-and-evidence needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Give an agent access to local evidence"],"missingBackbones":[],"stepCoverageGaps":["Give an agent access to local evidence > Connect a local server"],"rationale":"Prioritizes capabilities with delivery risk signals: interfaces/mcp/investigate-findings-and-evidence."}]},"presentation":{"outcome":"Read-only MCP server helps teams move from Connect a local server to Drill into findings and evidence across 1 backbone activity.","narrativePath":[{"backbone":"Give an agent access to local evidence","step":"Connect a local server","capabilityIds":["interfaces/mcp/connect-read-only-local-server"],"health":"review"},{"backbone":"Give an agent access to local evidence","step":"Drill into findings and evidence","capabilityIds":["interfaces/mcp/investigate-findings-and-evidence"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Give an agent access to local evidence > Connect a local server","message":"Give an agent access to local evidence > Connect a local server has needs-review coverage."},{"kind":"weak","label":"Give an agent access to local evidence > Drill into findings and evidence","message":"Give an agent access to local evidence > Drill into findings and evidence has needs-review coverage."}]}},{"release":"Related-metric analysis","capabilities":[{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/compare-declared-metrics.capability.yaml","health":"review","coverage":"partial","summary":"Explicit relationships produce ratio, difference, normalized-residual, and joint-condition series with source-level lineage.","intent":"Reveal changes between metrics that individual source values may not expose.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Derived series can be analyzed with the frozen or adaptive single-resolution recipe.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Differences, supplied-scale residuals, and all/any threshold conditions produce the declared derived values.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/compare-declared-metrics","title":"Find changes in conversion, unit cost, and metric gaps","criterion":"Missing sources and zero denominators remain unavailable with a reason rather than becoming zero.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/compare-declared-metrics --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","order":20}},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","status":"implemented","area":"relationships","path":".capabilities/relationships/comparisons/evaluate-delayed-response.capability.yaml","health":"review","coverage":"partial","summary":"A lagged-response relationship compares current response with a supplied or training-fitted multiple of an earlier predictor.","intent":"Inspect whether a metric still follows the delayed relationship declared by the caller.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"A supplied lag and coefficient use the predictor at the declared earlier sample position.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"Derived values retain response and predictor source indexes and source cutoffs in lineage.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/comparisons/evaluate-delayed-response","title":"Check whether a metric follows its expected delayed response","criterion":"Samples lacking the earlier predictor or awaiting coefficient training remain unavailable.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/comparisons/evaluate-delayed-response --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Check delayed responses","order":30}},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","status":"implemented","area":"relationships","path":".capabilities/relationships/quality/require-compatible-sources.capability.yaml","health":"review","coverage":"partial","summary":"Declared relationships require exact UTC timestamp alignment, shared cadence and entity scope, and compatible declared units.","intent":"Prevent comparisons between mismatched populations or periods from creating misleading evidence.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"An invalid relationship is reported as failed while valid sibling dataset results remain available.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"Difference relationships require identical declared source units and matching output units.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"relationships/quality/require-compatible-sources","title":"Compare the same entity and time period","criterion":"Related datasets use identical normalized timestamps, the same declared cadence, and matching entity metadata.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify relationships/quality/require-compatible-sources --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Related-metric analysis","backbone":"Investigate relationships I declare","step":"Align comparable sources","order":10}},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","status":"implemented","area":"relationships","path":".capabilities/relationships/resolution/compare-finalized-subperiods.capability.yaml","health":"review","coverage":"partial","summary":"Extend declared related-metric analysis to compatible finalized subperiods and completed fixed UTC 24-hour additive aggregates.","intent":"Detect conversion or unit-cost changes during a period while keeping partial progress separate from completed-period comparisons.","references":{"total":6,"readable":6,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":4,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":0,"ignored":0},"nextAction":"Run semantic review, save review evidence, or ignore accepted findings.","topFindings":[{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Callers select related datasets with identical native boundaries, entity scope, declared aggregation semantics, and compatible period definitions.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":19,"excerpt":"def view_cases(datasets, relationships, request):"},{"reference":"src/anomalyzer/sparse_views.py","line":22,"excerpt":"cases = compose_cases([d for d in datasets if d.resolution == resolution],"},{"reference":"src/anomalyzer/sparse_views.py","line":146,"excerpt":"native_sources = [d.model_copy(update={\"data_policy\": d.data_policy or DataPolicy()}) for d in request.datasets]"},{"reference":"src/anomalyzer/sparse_views.py","line":149,"excerpt":"native_request = request.model_copy(update={\"datasets\": native_sources, \"config\": native_config})"},{"reference":"src/anomalyzer/sparse_views.py","line":169,"excerpt":"daily_request = request.model_copy(update={\"datasets\": sources, \"config\": daily_config, \"relationships\": daily_relationships})"},{"reference":"src/anomalyzer/agent.py","line":14,"excerpt":"\"\"\"Use legacy semantics for a Dataset; explicit V11 fields select source policies.\"\"\""},{"reference":"src/anomalyzer/agent.py","line":20,"excerpt":"name in dataset for name in (\"data_policy\", \"coverage\", \"available_at\", \"revisions\", \"rules\", \"completed_period_rules\")):"},{"reference":"src/anomalyzer/agent.py","line":23,"excerpt":"return analyze({\"schema_version\": \"1.1\", \"datasets\": [validated.model_dump(mode=\"json\")],"},{"reference":"src/anomalyzer/agent.py","line":29,"excerpt":"return analyze({\"datasets\": [validated.model_dump(mode=\"json\")],"},{"reference":"tests/test_sparse.py","line":18,"excerpt":"return {\"schema_version\": \"1.1\", \"datasets\": [{\"id\": \"series\", \"entity\": {\"customer\": \"test\"},"},{"reference":"tests/test_sparse.py","line":60,"excerpt":"request[\"datasets\"][0][\"timestamps\"][2] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":88,"excerpt":"def test_declared_duplicate_aggregation_preserves_provenance(policy, value):"},{"reference":"tests/test_sparse.py","line":100,"excerpt":"request[\"datasets\"][0][\"timestamps\"][-1] = \"2026-01-01T03:00:00Z\""},{"reference":"tests/test_sparse.py","line":118,"excerpt":"changed[\"datasets\"][0][\"values\"][-10:] = [999]*10"},{"reference":"docs/sparse-data.md","line":35,"excerpt":"result = analyze({\"schema_version\": \"1.1\", \"datasets\": [source.model_dump(mode=\"json\")]})"},{"reference":"docs/sparse-data.md","line":136,"excerpt":"incomplete, excluded or reset boundaries, regardless of point-reference policy."},{"reference":"docs/sparse-data.md","line":142,"excerpt":"reset or caller-selected `reference_gaps=\"restart\"`; the default does not silently"},{"reference":"docs/sparse-data.md","line":172,"excerpt":"basis. A caller-declared maximum time separation can limit change comparisons."},{"reference":"docs/sparse-data.md","line":218,"excerpt":"Daily lagged, normalized-residual and joint-condition definitions remain"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Completed aggregate ratios use aggregated numerators and denominators; a ratio of totals is not silently replaced by an average of individual ratios.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":29,"excerpt":"case.non_claims.append(\"Native and completed-day views reuse the same observations; they are not independent confirmation.\")"},{"reference":"src/anomalyzer/sparse_views.py","line":70,"excerpt":"\"\"\"Sum only wholly observed finalized additive children. Never average ratios.\"\"\""},{"reference":"src/anomalyzer/sparse_views.py","line":83,"excerpt":"unsupported = \"completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged\""},{"reference":"src/anomalyzer/sparse_views.py","line":175,"excerpt":"applicability=\"completed-day view supports ratio-of-sums and differences of compatible additive sums only\","},{"reference":"src/anomalyzer/relationships.py","line":90,"excerpt":"used = [(numerator, index), (denominator, index)]"},{"reference":"src/anomalyzer/relationships.py","line":173,"excerpt":"\"complete\" if source.values[i] is not None else \"unknown_or_incomplete\" for source, i in used}"},{"reference":"tests/test_sparse.py","line":231,"excerpt":"request[\"relationships\"] = [{\"id\": \"unit-cost\", \"kind\": \"ratio\", \"numerator\": \"series\", \"denominator\": \"calls\", \"units\": \"dollars_per_call\"}]"},{"reference":"tests/test_sparse.py","line":260,"excerpt":"def test_daily_rules_are_not_silently_reused_from_hourly_rules():"},{"reference":"tests/test_sparse.py","line":279,"excerpt":"def test_adaptive_gap_does_not_silently_train_a_post_gap_episode():"},{"reference":"tests/test_sparse.py","line":322,"excerpt":"def test_restart_policy_cannot_reuse_a_prior_regime_reference():"},{"reference":"tests/test_sparse.py","line":431,"excerpt":"assert row[\"complete\"] and row[\"aggregate_value\"] == 24 and not row[\"model_eligible\"]"},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":120,"excerpt":"It then collects `calibration_size` eligible forecast residuals, not elapsed slots."},{"reference":"docs/sparse-data.md","line":142,"excerpt":"reset or caller-selected `reference_gaps=\"restart\"`; the default does not silently"},{"reference":"docs/sparse-data.md","line":213,"excerpt":"partial sums are progress metadata, never a completed total. Empty days are not"},{"reference":"docs/sparse-data.md","line":216,"excerpt":"Daily ratios divide summed compatible numerator and denominator quantities;"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."},{"capabilityId":"relationships/resolution/compare-finalized-subperiods","title":"Inspect related metrics before the business period closes","criterion":"Finalized subperiod relationships retain each source value and original source index; unavailable or incomplete sources yield explicit unavailability.","evidence":[{"reference":"src/anomalyzer/sparse_views.py","line":19,"excerpt":"def view_cases(datasets, relationships, request):"},{"reference":"src/anomalyzer/sparse_views.py","line":76,"excerpt":"unsupported = \"completed-day aggregation is unavailable on event_index\""},{"reference":"src/anomalyzer/sparse_views.py","line":95,"excerpt":"raise ValueError(\"source timestamps are off the aggregate_anchor grid\")"},{"reference":"src/anomalyzer/sparse_views.py","line":129,"excerpt":"source_input_indexes=[raw for i in indexes for raw in quality[\"samples\"][i][\"source_indexes\"]],"},{"reference":"src/anomalyzer/sparse_views.py","line":130,"excerpt":"source_grid_indexes=indexes, observed_sum=(math.fsum(measurements[i] for i in indexes if measurements[i] is not None)"},{"reference":"src/anomalyzer/relationships.py","line":7,"excerpt":"from .alignment import PreparedDataset, exact_sources"},{"reference":"src/anomalyzer/relationships.py","line":19,"excerpt":"def _source_ids(relationship: Relationship) -> list[str]:"},{"reference":"src/anomalyzer/relationships.py","line":33,"excerpt":"ids = _source_ids(relationship)"},{"reference":"src/anomalyzer/relationships.py","line":34,"excerpt":"sources = exact_sources(ids, prepared)"},{"reference":"src/anomalyzer/relationships.py","line":36,"excerpt":"sources = [replace(source, values=source.quality.get(\"rule_values\", source.values)) for source in sources]"},{"reference":"src/anomalyzer/agent.py","line":34,"excerpt":"def analyze_relationships(request: RequestV11 | dict) -> ResultV11:"},{"reference":"tests/test_sparse.py","line":10,"excerpt":"from anomalyzer import analyze, analyze_series, analyze_relationships, get_case, replay_policy"},{"reference":"tests/test_sparse.py","line":41,"excerpt":"assert right.data_quality[\"samples\"][8][\"source_index\"] == -1"},{"reference":"tests/test_sparse.py","line":51,"excerpt":"assert [r[\"original_value\"] for r in rows] == [2, None, None, 0, 9]"},{"reference":"tests/test_sparse.py","line":61,"excerpt":"assert [r[\"value\"] for r in source(request).data_quality[\"samples\"]] == [1, 0, 0, 2]"},{"reference":"tests/test_sparse.py","line":63,"excerpt":"assert [r[\"value\"] for r in source(request).data_quality[\"samples\"]] == [1, None, 0, 2]"},{"reference":"docs/sparse-data.md","line":21,"excerpt":"works through `analyze`, `analyze_relationships`, CLI JSON and read-only MCP."},{"reference":"docs/sparse-data.md","line":45,"excerpt":"| `coverage=\"incomplete\"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |"},{"reference":"docs/sparse-data.md","line":72,"excerpt":"Preparation reports original values, model values, coverage, source indexes,"},{"reference":"docs/sparse-data.md","line":76,"excerpt":"For policy-prepared sources, `grid_count` counts positions, `observation_count`"},{"reference":"docs/sparse-data.md","line":106,"excerpt":"retains its unique-source requirement. With an explicit `data_policy`, complete,"}],"status":"weak-evidence","confidence":"medium","action":"inspect-evidence","rationale":"Related implementation text was found, but deterministic matching cannot prove the behavior.","recommendation":"Review the cited evidence. If it is correct, save agent.review evidence or make the criterion/reference wording more explicit."}],"storyMap":{"release":"Related-metric analysis","backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","order":30}}],"deliveryStrategy":{"release":"Related-metric analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["relationships/quality/require-compatible-sources","relationships/resolution/compare-finalized-subperiods"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"relationships/quality/require-compatible-sources needs review; relationships/resolution/compare-finalized-subperiods needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Investigate relationships I declare","Monitor the periods my business uses"],"missingBackbones":[],"stepCoverageGaps":["Investigate relationships I declare > Check delayed responses","Investigate relationships I declare > Compare ratios and metric gaps"],"rationale":"Starts with one step from each mapped backbone: Investigate relationships I declare, Monitor the periods my business uses."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["relationships/comparisons/compare-declared-metrics","relationships/comparisons/evaluate-delayed-response"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"relationships/comparisons/compare-declared-metrics needs review; relationships/comparisons/evaluate-delayed-response needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Investigate relationships I declare"],"missingBackbones":["Monitor the periods my business uses"],"stepCoverageGaps":["Investigate relationships I declare > Align comparable sources","Monitor the periods my business uses > Compare related metrics while the period is open"],"rationale":"Prioritizes capabilities with delivery risk signals: relationships/comparisons/compare-declared-metrics, relationships/comparisons/evaluate-delayed-response."}]},"presentation":{"outcome":"Related-metric analysis helps teams move from Align comparable sources to Compare related metrics while the period is open across 2 backbone activities.","narrativePath":[{"backbone":"Investigate relationships I declare","step":"Align comparable sources","capabilityIds":["relationships/quality/require-compatible-sources"],"health":"review"},{"backbone":"Investigate relationships I declare","step":"Compare ratios and metric gaps","capabilityIds":["relationships/comparisons/compare-declared-metrics"],"health":"review"},{"backbone":"Investigate relationships I declare","step":"Check delayed responses","capabilityIds":["relationships/comparisons/evaluate-delayed-response"],"health":"review"},{"backbone":"Monitor the periods my business uses","step":"Compare related metrics while the period is open","capabilityIds":["relationships/resolution/compare-finalized-subperiods"],"health":"review"}],"coverageSignals":[{"kind":"missing","label":"Monitor the periods my business uses","message":"Monitor the periods my business uses is not covered in the recommended slice."},{"kind":"weak","label":"Investigate relationships I declare > Align comparable sources","message":"Investigate relationships I declare > Align comparable sources has needs-review coverage."},{"kind":"weak","label":"Investigate relationships I declare > Check delayed responses","message":"Investigate relationships I declare > Check delayed responses has needs-review coverage."},{"kind":"weak","label":"Investigate relationships I declare > Compare ratios and metric gaps","message":"Investigate relationships I declare > Compare ratios and metric gaps has needs-review coverage."},{"kind":"weak","label":"Monitor the periods my business uses > Compare related metrics while the period is open","message":"Monitor the periods my business uses > Compare related metrics while the period is open has needs-review coverage."}]}},{"release":"Trend and seasonal analysis","capabilities":[{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","status":"implemented","area":"analysis","path":".capabilities/analysis/detection/identify-residual-patterns.capability.yaml","health":"review","coverage":"partial","summary":"Residual rules identify location shifts, trends, oscillation, mixture patterns, and short-term variation beyond isolated point departures.","intent":"Find accumulating changes and explain which numerical pattern was detected.","references":{"total":3,"readable":3,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"CUSUM reports the first detection position, direction, accumulated score, and configured threshold.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"Moving-range findings identify variation changes separately from location changes.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/detection/identify-residual-patterns","title":"Spot sustained shifts and changing patterns","criterion":"Nelson location rules and CUSUM can identify a sustained residual location shift without any point-threshold crossing.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/detection/identify-residual-patterns --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","order":60}},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/adapt-to-changing-growth.capability.yaml","health":"review","coverage":"partial","summary":"The adaptive recipe updates each expectation from a bounded window of earlier observations while retaining frozen residual calibration.","intent":"Follow evolving growth without losing evidence of an isolated unusual value.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"An isolated extreme value is held out of future model references rather than creating a seasonal echo.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"Conflicting trend choices and slope lookbacks exceeding the adaptive window are rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/adapt-to-changing-growth","title":"Follow evolving growth while exposing unusual values","criterion":"Each rolling expectation uses only observations before the sample being scored.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/adapt-to-changing-growth --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","order":40}},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/model-seasonal-level-and-growth.capability.yaml","health":"review","coverage":"partial","summary":"The default recipe builds an expectation from earlier matching seasonal observations and a training-fitted level, linear, or compound-growth model.","intent":"Distinguish unusual departures from ordinary repeating patterns and growth.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"A supplied season length selects earlier matching seasonal observations for the reference.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"Automatic trend selection represents suitable linear and compound growth without flagging ordinary growth as point anomalies.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/model-seasonal-level-and-growth","title":"Account for ordinary seasonal patterns and growth","criterion":"Callers can explicitly select none, linear, or exponential trend, and unsupported trend modes are rejected.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/model-seasonal-level-and-growth --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","order":10}},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/protect-reference-from-extremes.capability.yaml","health":"review","coverage":"partial","summary":"Robust mode limits extreme observations' influence on trend fitting, residual calibration, and later seasonal references.","intent":"Keep one incident from inflating the reference or appearing again as a seasonal echo.","references":{"total":7,"readable":7,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"An extreme evaluated sample remains evidence while its expected value is used in future references.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"Callers can choose include mode to retain raw extreme observations and inspect the resulting echo behavior.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/protect-reference-from-extremes","title":"Limit how extreme values distort later comparisons","criterion":"Robust seasonal references reduce the influence of an extreme prior matching seasonal value.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/protect-reference-from-extremes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","order":30}},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/reset-declared-regimes.capability.yaml","health":"review","coverage":"partial","summary":"Caller-supplied reset points divide a single-resolution series into independently trained and calibrated regimes.","intent":"Recognize known discontinuities without forcing the previous regime to explain the new one.","references":{"total":4,"readable":4,"missing":[]},"verification":{"automated":3,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"CLI reset options produce the same resolved positions and suppress departures explained by the declared boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"Each new regime rebuilds its reference, trend, and calibration without crossing earlier boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/reset-declared-regimes","title":"Compare against a new baseline after a known change","criterion":"Integer positions and exact observed dates or timestamps can identify reset boundaries.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/reset-declared-regimes --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","order":50}},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","status":"implemented","area":"analysis","path":".capabilities/analysis/expectations/select-season-length.capability.yaml","health":"review","coverage":"partial","summary":"Callers can supply a season length; ordered values without one can use conservative season inference with a lag-one fallback.","intent":"Make seasonal assumptions explicit and inspectable for data without timestamps.","references":{"total":5,"readable":5,"missing":[]},"verification":{"automated":4,"manual":1,"gaps":[]},"counts":{"covered":0,"weak-evidence":0,"implementation-gap":0,"missing-reference":0,"no-implementation-reference":0,"assessor-limitation":4,"ignored":0},"nextAction":"Split broad criteria or run semantic review.","topFindings":[{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"A positional series without a sufficiently stable candidate falls back to lag one and records why.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"A repeating positional series can infer a seasonal lag and report the selection window.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]},{"capabilityId":"analysis/expectations/select-season-length","title":"Choose the repeating cycle for a fair comparison","criterion":"An explicit season length is honored and reported as explicit.","status":"assessor-limitation","confidence":"low","action":"manual-review","rationale":"Saved review is stale: capability or implementation evidence changed.","recommendation":"Run capabilitykit verify analysis/expectations/select-season-length --agent <command> to refresh this capability.","evidence":[]}],"storyMap":{"release":"Trend and seasonal analysis","backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","order":20}}],"deliveryStrategy":{"release":"Trend and seasonal analysis","recommendations":[{"order":1,"phase":"opening","name":"Walking skeleton","capabilityIds":["analysis/expectations/model-seasonal-level-and-growth"],"releaseStrategy":"Prove one coherent end-to-end slice before deepening individual steps.","developmentStrategy":"Integrate the earliest capability from each backbone first, keeping scope thin until the narrative works.","riskIntent":"analysis/expectations/model-seasonal-level-and-growth needs review","learningIntent":"Validate whether users and stakeholders can recognize a meaningful outcome across the release backbone.","backboneCoverage":["Refine what ordinary behavior means"],"missingBackbones":[],"stepCoverageGaps":["Refine what ordinary behavior means > Choose a repeating cycle","Refine what ordinary behavior means > Find sustained residual changes","Refine what ordinary behavior means > Follow evolving growth","Refine what ordinary behavior means > Protect later comparisons","Refine what ordinary behavior means > Restart after a reviewed change"],"rationale":"Starts with the earliest mapped step in Refine what ordinary behavior means."},{"order":2,"phase":"end-game","name":"Risk burn-down and release hardening","capabilityIds":["analysis/expectations/select-season-length","analysis/expectations/protect-reference-from-extremes","analysis/expectations/adapt-to-changing-growth","analysis/expectations/reset-declared-regimes","analysis/detection/identify-residual-patterns"],"releaseStrategy":"Resolve delivery risks once the release story is visible enough to judge tradeoffs.","developmentStrategy":"Target review, action, verification, and reference gaps with explicit finish-line criteria.","riskIntent":"analysis/detection/identify-residual-patterns needs review; analysis/expectations/adapt-to-changing-growth needs review; analysis/expectations/protect-reference-from-extremes needs review; analysis/expectations/reset-declared-regimes needs review; analysis/expectations/select-season-length needs review","learningIntent":"Learn whether remaining risks change scope, release confidence, or stakeholder messaging.","backboneCoverage":["Refine what ordinary behavior means"],"missingBackbones":[],"stepCoverageGaps":["Refine what ordinary behavior means > Account for growth and seasonality"],"rationale":"Prioritizes capabilities with delivery risk signals: analysis/expectations/select-season-length, analysis/expectations/protect-reference-from-extremes, analysis/expectations/adapt-to-changing-growth, analysis/expectations/reset-declared-regimes, analysis/detection/identify-residual-patterns."}]},"presentation":{"outcome":"Trend and seasonal analysis helps teams move from Account for growth and seasonality to Find sustained residual changes across 1 backbone activity.","narrativePath":[{"backbone":"Refine what ordinary behavior means","step":"Account for growth and seasonality","capabilityIds":["analysis/expectations/model-seasonal-level-and-growth"],"health":"review"},{"backbone":"Refine what ordinary behavior means","step":"Choose a repeating cycle","capabilityIds":["analysis/expectations/select-season-length"],"health":"review"},{"backbone":"Refine what ordinary behavior means","step":"Protect later comparisons","capabilityIds":["analysis/expectations/protect-reference-from-extremes"],"health":"review"},{"backbone":"Refine what ordinary behavior means","step":"Follow evolving growth","capabilityIds":["analysis/expectations/adapt-to-changing-growth"],"health":"review"},{"backbone":"Refine what ordinary behavior means","step":"Restart after a reviewed change","capabilityIds":["analysis/expectations/reset-declared-regimes"],"health":"review"},{"backbone":"Refine what ordinary behavior means","step":"Find sustained residual changes","capabilityIds":["analysis/detection/identify-residual-patterns"],"health":"review"}],"coverageSignals":[{"kind":"weak","label":"Refine what ordinary behavior means > Account for growth and seasonality","message":"Refine what ordinary behavior means > Account for growth and seasonality has needs-review coverage."},{"kind":"weak","label":"Refine what ordinary behavior means > Choose a repeating cycle","message":"Refine what ordinary behavior means > Choose a repeating cycle has needs-review coverage."},{"kind":"weak","label":"Refine what ordinary behavior means > Find sustained residual changes","message":"Refine what ordinary behavior means > Find sustained residual changes has needs-review coverage."},{"kind":"weak","label":"Refine what ordinary behavior means > Follow evolving growth","message":"Refine what ordinary behavior means > Follow evolving growth has needs-review coverage."},{"kind":"weak","label":"Refine what ordinary behavior means > Protect later comparisons","message":"Refine what ordinary behavior means > Protect later comparisons has needs-review coverage."},{"kind":"weak","label":"Refine what ordinary behavior means > Restart after a reviewed change","message":"Refine what ordinary behavior means > Restart after a reviewed change has needs-review coverage."}]}}]},"summary":{"total":48,"ok":0,"review":30,"action":1,"planned":17}};
 let selectedRelease = "";
 let selectedHealth = "";
 let searchText = "";
diff --git a/.github/workflows/capability-maps.yml b/.github/workflows/capability-maps.yml
index 749d8df..59fae45 100644
--- a/.github/workflows/capability-maps.yml
+++ b/.github/workflows/capability-maps.yml
@@ -6,12 +6,22 @@ on:
     paths:
       - '.capabilities/**'
       - 'docs/capability-maps/**'
+      - 'README.md'
+      - 'docs/sparse-data.md'
+      - 'docs/images/**'
+      - 'examples/sparse/**'
+      - 'scripts/export_examples_site.py'
       - 'scripts/export_capability_preview.py'
       - '.github/workflows/capability-maps.yml'
   pull_request:
     paths:
       - '.capabilities/**'
       - 'docs/capability-maps/**'
+      - 'README.md'
+      - 'docs/sparse-data.md'
+      - 'docs/images/**'
+      - 'examples/sparse/**'
+      - 'scripts/export_examples_site.py'
       - 'scripts/export_capability_preview.py'
       - '.github/workflows/capability-maps.yml'
   workflow_dispatch:
@@ -33,6 +43,9 @@ jobs:
           python-version: '3.13'
       - name: Check static preview
         run: python scripts/export_capability_preview.py --check
+      - uses: astral-sh/setup-uv@v6
+        with:
+          version: '0.6.13'
       - name: Assemble public map files
         run: |
           mkdir -p _site
@@ -40,6 +53,8 @@ jobs:
           cp .capabilities/dependency-viewer.html _site/dependency-viewer.html
           cp .capabilities/story-map-viewer.html _site/story-map-viewer.html
           cp .capabilities/dependency-preview.svg _site/dependency-preview.svg
+      - name: Assemble numbered examples and API guide
+        run: uv run --locked --extra docs python scripts/export_examples_site.py --output _site/examples
       - uses: actions/upload-pages-artifact@v4
         with:
           path: _site
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8bf2bd1..2451c8e 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -26,6 +26,8 @@ jobs:
           enable-cache: true
           python-version: ${{ matrix.python-version }}
       - run: uv sync --locked --extra dev
+      - run: uv run python examples/generate.py
+      - run: git diff --exit-code -- examples
       - run: uv run pytest -q
       - run: uv run python examples/run_demo.py
       - run: uv run python examples/run_agentic_demo.py
@@ -105,3 +107,4 @@ jobs:
             .tmp/lambda-consumer/smoke.json
             .tmp/lambda-consumer/requirements-offline.txt
             .tmp/lambda-consumer/wheelhouse-SHA256SUMS
+            .tmp/lambda-consumer/reports/benchmark.json
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 890bd74..99a154c 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -3,6 +3,8 @@ name: Release
 on:
   push:
     tags: ['v*']
+  release:
+    types: [published]
 
 permissions:
   contents: read
@@ -13,11 +15,12 @@ concurrency:
 
 jobs:
   checks:
+    if: github.event_name == 'push'
     uses: ./.github/workflows/ci.yml
 
   publish:
     needs: checks
-    if: github.repository == 'PredictabilityAtScale/anomaly-detection'
+    if: github.event_name == 'push' && github.repository == 'PredictabilityAtScale/anomaly-detection'
     runs-on: ubuntu-latest
     permissions:
       contents: write
@@ -35,3 +38,26 @@ jobs:
           GH_TOKEN: ${{ github.token }}
           RELEASE_TAG: ${{ github.ref_name }}
         run: python scripts/publish_release.py --tag "$RELEASE_TAG" --revision "$GITHUB_SHA"
+
+  verify-published:
+    if: github.event_name == 'release'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event.release.tag_name }}
+      - uses: astral-sh/setup-uv@v6
+        with:
+          version: '0.6.13'
+          python-version: '3.12'
+      - name: Download and verify the immutable published assets
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: ${{ github.event.release.tag_name }}
+        run: |
+          gh release download "$RELEASE_TAG" --repo "$GITHUB_REPOSITORY" --dir dist
+          python scripts/release.py verify --tag "$RELEASE_TAG" --revision "$(git rev-parse HEAD)"
+      - name: Install the exact published wheel URL with hashes outside checkout
+        run: python scripts/verify_distribution.py --python 3.12 --from-release
+      - name: Verify the downloaded release in the offline Lambda-compatible bundle
+        run: bash scripts/verify_lambda.sh
diff --git a/README.md b/README.md
index 9c6a191..6601e7f 100644
--- a/README.md
+++ b/README.md
@@ -192,7 +192,8 @@ when visualization is unavailable.
 Schema 1.0 is the single-series contract. Schema 1.1 adds named datasets,
 relationships, assessments, policy, and cases; it also works with one dataset
 and no relationships. Both use the same analysis core through CLI, Python, and
-MCP. The multi-resolution recipe currently supports schema 1.0 only and does
+MCP. Sparse schema-1.1 and multi-resolution support is described in chapters 14–22.
+The legacy schema-1.0 multi-resolution profile does
 not support `reset_points`; both other recipes support explicit resets.
 
 ## Capability maps
@@ -597,6 +598,137 @@ reanalysis. Older tool names remain callable for compatibility but are not
 advertised. Recurring monitoring, deduplication of previously reported findings,
 review state, alerts, and production actions require caller-held state and logic.
 
+### 14. Choose what an empty period means
+
+The same timestamps can mean different things when collection guarantees differ.
+The safe default preserves missing values. An explicit additive-volume policy can
+interpret **confirmed complete empty periods** as zero. Unknown collection never
+provides that guarantee. A typical value, expectation, interpolation or previous
+measurement is never inserted as an observation.
+
+![Same input with preserved gaps versus explicitly authorized zero volume](docs/images/anomaly-progression-14-empty-policy.png)
+
+Compare [preserve.json](examples/sparse/preserve.json) with
+[zero_empty.json](examples/sparse/zero_empty.json). The latter declares
+`data_policy={"measurement":"additive_volume","missing":"zero_for_empty"}`;
+its four `coverage="empty"` positions have visible conversion reasons. The blue
+observations are unchanged; gold markers distinguish policy-derived zeros.
+
+### 15. Keep zero, unknown and unfinished distinct
+
+An observed zero is a real value. An unknown period stays a gap. An open period
+can show its reported progress without entering a completed baseline or total.
+The [coverage example](examples/sparse/statuses.json) retains the original values,
+statuses and grid, including the open period's value of 3.
+
+![Observed zeros, unknown gaps and an unfinished value shown separately](docs/images/anomaly-progression-15-coverage-states.png)
+
+Missing periods cannot establish normality or recovery. An expectation may remain
+visible as a separate dashed reference across a gap; the observed line breaks.
+
+### 16. Retain a quiet customer's later event
+
+For a completely collected additive count, quiet periods can legitimately be
+zero. [This synthetic customer](examples/sparse/quiet_volume.json) normally has one
+call every 12 hours. Confirmed quiet hours remain in their actual seasonal phase;
+a later measurement of 40 calls is still assessable.
+
+![A quiet but completely collected volume series and a later large measurement](docs/images/anomaly-progression-16-quiet-volume.png)
+
+The profile learns from eligible measurements and authorized empty counts.
+Business materiality, severity and incident decisions remain the caller's job.
+
+### 17. Compare real calls across quiet hours
+
+No calls means cost per call is undefined. The [calendar profile](examples/sparse/cost_calendar.json)
+keeps 431 hourly positions, but lacks matching daily-phase support. The
+[event-index profile](examples/sparse/cost_events.json) explicitly compares 12
+genuine comparable calls and retains every call's original event timestamp.
+
+![Undefined quiet-hour cost values versus comparable event-index observations](docs/images/anomaly-progression-17-event-index.png)
+
+Set `data_policy.observation_basis="event_index"`, omit `frequency`, and use
+`season_length=1`. This is event position, not elapsed time or a calendar rate.
+Model/workload compatibility is a caller declaration; tokens or consumed units
+are not implicitly independent events.
+
+### 18. Keep the weekday when observations are absent
+
+Sparse calendar history must not be squeezed into a shorter positional sequence.
+The [weekday example](examples/sparse/weekday_phase.json) preserves five missing
+days. Its reference still compares the same seven-day phase, using eligible
+training values and a fixed calibration origin.
+
+![Sparse observations retaining their actual seven-day seasonal phase](docs/images/anomaly-progression-18-sparse-phase.png)
+
+The sparse frozen profile can borrow supported training phases across gaps.
+Unsupported phases remain unavailable. Phase support, calibration eligibility and
+reference quality are reported separately; evaluation samples do not update this
+frozen reference. Adaptive and explicit trend recipes currently need complete
+contiguous spans and an explicit reset/restart before new post-gap learning.
+
+### 19. Restart consecutive patterns at a gap
+
+Point-reference reuse does not make missing residuals consecutive. In
+[this example](examples/sparse/pattern_gap.json), four positive residuals, one
+unknown hour and seven positive residuals do not make an eleven-observation run.
+CUSUM restarts independently; Nelson Rule 2 does not bridge the gap.
+
+![Consecutive residual patterns broken by an unknown collection period](docs/images/anomaly-progression-19-pattern-gap.png)
+
+Unknown, incomplete and explicitly excluded model samples break pattern runs.
+Reviewed resets also break early comparisons, references and calibration.
+
+### 20. Apply a measured rule on the first observation
+
+A caller-supplied range can be checked before any statistical learning. The
+[first-value example](examples/sparse/first_rule.json) observes 100 measured units
+against an explicitly declared acceptable range of 0–20. It retains the rule
+assessment and case while statistical status is `insufficient_history`.
+
+![First-observation numerical rule violation while statistical learning is unavailable](docs/images/anomaly-progression-20-first-rule.png)
+
+This threshold is illustrative, not a recommended business-risk limit. Change
+rules need a comparable prior observation and respect gaps/resets; relative
+change from zero stays unavailable. Early descriptive/provisional evidence alone
+cannot trigger a case. Numerical policy eligibility is not operational authority.
+
+### 21. Keep undefined ratios unavailable
+
+[Cost per call](examples/sparse/relationships.json) requires an aligned valid
+numerator and denominator. At hour 1 the denominator is zero; hours 2 and 9 have
+missing inputs. These points stay undefined with explicit reasons and source
+lineage, even when another source uses an additive zero policy.
+
+![Aligned source quantities and undefined derived ratios shown with real gaps](docs/images/anomaly-progression-21-undefined-ratios.png)
+
+An observed numerator of zero over positive volume can produce an observed ratio
+of zero. A missing numerator or zero denominator cannot.
+
+### 22. Preserve incomplete days and useful hourly evidence
+
+A daily total needs every expected finalized child. In the
+[sparse daily example](examples/sparse/unfinished_daily.json), day 1 has an unknown
+hour, day 2 is complete, and day 3 remains open. Only day 2 supplies a completed
+total. Eligible hourly evidence remains available while daily learning lacks
+support.
+
+![Unknown child and unfinished day kept out of completed totals while hourly evidence survives](docs/images/anomaly-progression-22-sparse-days.png)
+
+Schema-1.1 multi-resolution outputs label each view and retain assessments,
+relationships, cases and policy replay. Daily ratios divide summed compatible
+quantities; they do not average percentages. Daily thresholds belong in separate
+`completed_period_rules`. Views are correlated evidence, not independent votes.
+This supports fixed UTC days and frozen seasonal children; local-calendar/DST
+aggregation, adaptive multi-resolution children and simultaneous multi-seasonality
+remain separate work.
+
+The [sparse-data API guide](docs/sparse-data.md) documents policy defaults, bounded
+grid expansion, event-index restrictions, availability/correction semantics and
+the remaining complete-span restrictions. All new fixtures run through
+`examples/run_demo.py`; the existing generator and figure renderer reproduce the
+full progression.
+
 ## Baseline and calibration reference
 
 The robust expectation for sample `t` uses four previous matching seasonal
@@ -1010,7 +1142,7 @@ local DST times require offsets. Cadence is a fixed duration, such as `1h`, `1d`
 `1w`, or `15min`; days are 24 elapsed hours and weeks are seven such days.
 Sorting and UTC normalization are
 reported. Duplicates require `--duplicate-policy mean|sum`; nulls propagate.
-Missing values or irregular cadence make the baseline inapplicable. No imputation,
+In schema 1.0, missing values or irregular cadence make the baseline inapplicable. No implicit imputation,
 resampling, interpolation, or unit conversion occurs.
 
 `--config` accepts settings or a complete canonical request. Precedence is
diff --git a/docs/capability-maps/index.html b/docs/capability-maps/index.html
index 3862e02..bdda292 100644
--- a/docs/capability-maps/index.html
+++ b/docs/capability-maps/index.html
@@ -18,10 +18,12 @@
   <main>
     <h1>Anomalyzer capability maps</h1>
     <p>What changed, compared with what should have happened, and what can we conclude?</p>
-    <p>Start with the story map to explore eight current deliverable lanes and six
-      future lanes containing 18 planned stories. Open a capability
+    <p>Start with the story map to explore current deliverable lanes and
+      future work with explicitly planned stories. Open a capability
       to inspect its behavior, acceptance criteria, supporting evidence, and limits.</p>
     <ul>
+      <li><a href="examples/">Numbered feature examples, chapters 1–22</a> —
+        inspect real synthetic results, chart gaps and the sparse-data API guide.</li>
       <li><a href="story-map-viewer.html">Interactive deliverable story map</a> —
         choose an analysis extension or a way to use the product.</li>
       <li><a href="dependency-viewer.html">Interactive dependency map</a> —
@@ -39,7 +41,7 @@ <h2>Current deliverables</h2>
       <li>Read-only MCP server: let an agent inspect local findings and evidence.</li>
     </ol>
     <h2>Future direction — planned, unimplemented</h2>
-    <p>Each future lane has three ordered stories. Their acceptance criteria and
+    <p>Future lanes contain ordered planned stories. Their acceptance criteria and
       proposed verification scenarios describe intended behavior; implementation
       and review evidence will be added as the work is delivered.</p>
     <ol>
diff --git a/docs/images/anomaly-progression-14-empty-policy.png b/docs/images/anomaly-progression-14-empty-policy.png
new file mode 100644
index 0000000..fa1a866
Binary files /dev/null and b/docs/images/anomaly-progression-14-empty-policy.png differ
diff --git a/docs/images/anomaly-progression-15-coverage-states.png b/docs/images/anomaly-progression-15-coverage-states.png
new file mode 100644
index 0000000..6b65116
Binary files /dev/null and b/docs/images/anomaly-progression-15-coverage-states.png differ
diff --git a/docs/images/anomaly-progression-16-quiet-volume.png b/docs/images/anomaly-progression-16-quiet-volume.png
new file mode 100644
index 0000000..43298df
Binary files /dev/null and b/docs/images/anomaly-progression-16-quiet-volume.png differ
diff --git a/docs/images/anomaly-progression-17-event-index.png b/docs/images/anomaly-progression-17-event-index.png
new file mode 100644
index 0000000..6680ebb
Binary files /dev/null and b/docs/images/anomaly-progression-17-event-index.png differ
diff --git a/docs/images/anomaly-progression-18-sparse-phase.png b/docs/images/anomaly-progression-18-sparse-phase.png
new file mode 100644
index 0000000..204a454
Binary files /dev/null and b/docs/images/anomaly-progression-18-sparse-phase.png differ
diff --git a/docs/images/anomaly-progression-19-pattern-gap.png b/docs/images/anomaly-progression-19-pattern-gap.png
new file mode 100644
index 0000000..c97d002
Binary files /dev/null and b/docs/images/anomaly-progression-19-pattern-gap.png differ
diff --git a/docs/images/anomaly-progression-20-first-rule.png b/docs/images/anomaly-progression-20-first-rule.png
new file mode 100644
index 0000000..e524559
Binary files /dev/null and b/docs/images/anomaly-progression-20-first-rule.png differ
diff --git a/docs/images/anomaly-progression-21-undefined-ratios.png b/docs/images/anomaly-progression-21-undefined-ratios.png
new file mode 100644
index 0000000..58e5a06
Binary files /dev/null and b/docs/images/anomaly-progression-21-undefined-ratios.png differ
diff --git a/docs/images/anomaly-progression-22-sparse-days.png b/docs/images/anomaly-progression-22-sparse-days.png
new file mode 100644
index 0000000..b31f8a3
Binary files /dev/null and b/docs/images/anomaly-progression-22-sparse-days.png differ
diff --git a/docs/releases/0.2.0.md b/docs/releases/0.2.0.md
new file mode 100644
index 0000000..2ec1e23
--- /dev/null
+++ b/docs/releases/0.2.0.md
@@ -0,0 +1,38 @@
+# Anomalyzer 0.2.0 — explicit sparse-data semantics
+
+Applications can preserve unknown calendar periods, declare confirmed empty
+additive volume, or use comparable event-index measurements without inventing
+regular timestamps. Schema 1.0 remains unchanged; schema 1.1 gains per-source
+policies, coverage, availability/revision inputs and inspectable transformations.
+
+- Frozen sparse references retain actual calendar phase, count eligible
+  training/calibration observations and report unsupported phases.
+- Missing/incomplete/excluded positions and resets break patterns and change
+  comparisons. A measured range rule can operate before statistical learning.
+- Relationships preserve unavailable reasons, zero-denominator semantics,
+  original source lineage and model exclusions.
+- Sparse schema-1.1 multi-resolution retains native evidence and completed UTC
+  day views. Daily ratios divide sums; incomplete child periods cannot create a
+  completed total. Daily rules are explicit and cases survive policy replay.
+- README progression chapters 14–22 include reproducible fixtures, checked
+  results and rendered figures, published with the existing documentation site.
+
+See the sparse-data API guide at `docs/sparse-data.md` and the release/consumer
+guide at `docs/releases/README.md`. The installed-artifact checks cover the full
+public API and an offline Linux x86_64/Python 3.12 Lambda-compatible bundle.
+CI also records bounded 1,344-hour and event-index benchmark results and result
+sizes. These are synthetic mechanics/performance checks, not abuse accuracy or
+production cost claims.
+
+Explicit limits: event-index timestamps must be unique; adaptive/explicit trend
+recipes use complete contiguous spans; multi-resolution children are frozen and
+use fixed UTC days. Daily lagged/normalized/joint relationships, local-calendar
+DST aggregation, simultaneous multi-seasonality, episode quarantine and durable
+incremental state/diffing remain separate work.
+
+Guard continues to own collection contracts, profile compatibility, persistence,
+scheduling, business impact/severity, incident lifecycle and opt-in notifications.
+This release does not modify or deploy UsageTap Guard. Install the exact release
+wheel and hash-pinned dependencies during the application's Linux build and
+bundle them for offline runtime; no service, checkout, CLI or MCP process is
+required. PyPI publication is not part of this release.
diff --git a/docs/releases/README.md b/docs/releases/README.md
index 1726792..a2e0bb9 100644
--- a/docs/releases/README.md
+++ b/docs/releases/README.md
@@ -1,6 +1,6 @@
 # Release and use Anomalyzer
 
-Anomalyzer is a normal Python dependency. Version **0.1.0** is the first release
+Anomalyzer is a normal Python dependency. Version **0.2.0** is the first release
 candidate; nothing is uploaded to PyPI by this workflow. GitHub Releases is the
 initial distribution channel. Publishing the library does not integrate it into
 UsageTap Guard.
@@ -11,7 +11,7 @@ UsageTap Guard.
 | --- | --- |
 | Wheel (`.whl`) | Ready-to-install Python package. Use this for Guard. Anomalyzer's wheel works across supported operating systems; its dependencies may need platform-specific wheels. |
 | Source distribution (`.tar.gz`, or sdist) | Source needed to build the package. Useful for rebuilding and inspection; installation also needs the declared build backend. |
-| Version tag (`v0.1.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |
+| Version tag (`v0.2.0`) | Permanent Git name for the exact source revision released. It must match `_version.py`; never move or reuse it. |
 | Dependency pin | An exact version or artifact URL, plus a SHA-256 fingerprint. A hash checks that downloaded bytes match the intended artifact. Pin transitive dependencies too. |
 
 The package supports Python 3.11, 3.12 and 3.13. Version lives only in
@@ -38,10 +38,10 @@ it for package metadata and the public `__version__` imports it.
    and sign into GitHub CLI with your own account (`gh auth login`) if you use
    the optional CLI commands. Never paste credentials into this repository.
 
-At preparation time there were no existing tags/releases and release
-immutability was disabled. Check again before choosing a version.
+Before choosing a version, inspect existing tags/releases and the immutability
+setting. Never reuse a published version or move a version tag.
 
-## First release: 0.1.0
+## First release: 0.2.0
 
 Run these commands from a clean checkout of the merged `main` branch. Commands
 work in PowerShell or Bash except where explicitly labelled.
@@ -57,16 +57,16 @@ uv sync --locked --extra dev
 uv run pytest -q
 uv run python examples/run_demo.py
 uv run python examples/run_agentic_demo.py
-uv run python scripts/release.py check-version --tag v0.1.0
+uv run python scripts/release.py check-version --tag v0.2.0
 ```
 
-`git status --short` must be empty. If `v0.1.0` already exists, stop and choose
+`git status --short` must be empty. If `v0.2.0` already exists, stop and choose
 the next unused version through a PR. Otherwise create an annotated tag on the
 reviewed commit and push it (do not use `--force`):
 
 ```text
-git tag -a v0.1.0 -m "Anomalyzer 0.1.0"
-git push origin v0.1.0
+git tag -a v0.2.0 -m "Anomalyzer 0.2.0"
+git push origin v0.2.0
 ```
 
 Open the repository's **Actions → Release** run. It runs all tests on
@@ -78,10 +78,10 @@ passes does it create a draft and attach the tested files. Publication is one
 explicit owner step, using the already tested files:
 
 ```text
-git switch --detach v0.1.0
-gh release download v0.1.0 --repo PredictabilityAtScale/anomaly-detection --dir release-v0.1.0
+git switch --detach v0.2.0
+gh release download v0.2.0 --repo PredictabilityAtScale/anomaly-detection --dir release-v0.2.0
 git rev-parse HEAD
-python scripts/publish_release.py --dist release-v0.1.0 --tag v0.1.0 --revision <full-SHA-printed-above> --publish
+python scripts/publish_release.py --dist release-v0.2.0 --tag v0.2.0 --revision <full-SHA-printed-above> --publish
 git switch main
 ```
 
@@ -91,18 +91,23 @@ the actual repository immutability setting, remote tag, checksums and existing
 draft bytes before publishing. It never rebuilds. GitHub then seals the tag and
 all assets together. If GitHub CLI reports a permission problem, use the owner
 account (`gh auth switch`) or sign in as that owner; do not add a token to CI.
+Publication also triggers `verify-published`: it downloads and checks the sealed
+assets, installs the exact public wheel URL with its hashes in a clean consumer,
+and repeats the offline Linux/Python 3.12 bundle test. Check that job before
+adopting the release. Transient download failures can be rerun without changing
+any release asset.
 
 Assets live under [GitHub Releases](https://github.com/PredictabilityAtScale/anomaly-detection/releases),
 in the release for the selected tag:
 
-- `anomalyzer-0.1.0-py3-none-any.whl` and `anomalyzer-0.1.0.tar.gz`.
+- `anomalyzer-0.2.0-py3-none-any.whl` and `anomalyzer-0.2.0.tar.gz`.
 - `release-manifest.json`: version, tag, full source commit, file sizes and hashes.
 - `SHA256SUMS`: hashes of all assets including the manifest (except itself).
 - `runtime-dependencies.txt`: exact dependency versions and accepted hashes
   exported from the repository lock, including platform markers.
 - `requirements-guard-py312-linux-x86_64.txt`: those dependencies plus the exact
   GitHub wheel URL and its hash. This is an installable starting lock for Guard.
-- `requirements-offline.txt`: the same pins using `anomalyzer==0.1.0` so pip can
+- `requirements-offline.txt`: the same pins using `anomalyzer==0.2.0` so pip can
   install entirely from a local wheelhouse.
 
 CI's **Artifacts** section also holds `release-dist-<commit>` and the
@@ -147,11 +152,11 @@ analysis needs no checkout, CLI, MCP server, hosted Anomalyzer service or networ
 1. Download the chosen Release's assets and verify `SHA256SUMS`. For example:
 
    ```text
-   gh release download v0.1.0 --repo PredictabilityAtScale/anomaly-detection --dir release
+   gh release download v0.2.0 --repo PredictabilityAtScale/anomaly-detection --dir release
    ```
 
    In Linux/WSL/Bash: `(cd release && sha256sum --check SHA256SUMS)`.
-   In PowerShell, `Get-FileHash release/anomalyzer-0.1.0-py3-none-any.whl -Algorithm SHA256`
+   In PowerShell, `Get-FileHash release/anomalyzer-0.2.0-py3-none-any.whl -Algorithm SHA256`
    shows the wheel's hash; compare it with the manifest/checksum file. Also check
    the manifest's tag and commit against the intended release. Hashes identify
    bytes; use the trusted GitHub repository/release as their source.
@@ -160,7 +165,7 @@ analysis needs no checkout, CLI, MCP server, hosted Anomalyzer service or networ
    inputs. Its final requirement has this form, with a real generated hash:
 
    ```text
-   anomalyzer @ https://github.com/PredictabilityAtScale/anomaly-detection/releases/download/v0.1.0/anomalyzer-0.1.0-py3-none-any.whl --hash=sha256:<generated-wheel-hash>
+   anomalyzer @ https://github.com/PredictabilityAtScale/anomaly-detection/releases/download/v0.2.0/anomalyzer-0.2.0-py3-none-any.whl --hash=sha256:<generated-wheel-hash>
    ```
 
    Install the actual file, not this placeholder. It also pins Pydantic,
diff --git a/docs/sparse-data.md b/docs/sparse-data.md
new file mode 100644
index 0000000..46d52ea
--- /dev/null
+++ b/docs/sparse-data.md
@@ -0,0 +1,260 @@
+# Sparse observations in Anomalyzer 0.2
+
+An empty hour does not tell us whether nothing happened or collection failed.
+Schema 1.1 therefore preserves missing values by default. An expectation can be
+shown separately; it is never inserted as an observed measurement. Schema 1.0
+keeps its existing complete-series behavior and unchanged schemas.
+Schema 1.1 gains optional source and result fields. Upgrade strict schema-1.1
+producers/readers together to 0.2; an older strict result model is not promised to
+accept the new metadata. Complete legacy requests remain accepted. Record the
+package version and method `profile_revision` with Guard's own profile revision.
+
+The numbered [README progression](../README.md#14-choose-what-an-empty-period-means)
+continues with chapters 14–22. Fixtures live in [examples/sparse](../examples/sparse/).
+Run `python examples/generate.py`, `python examples/run_demo.py` and
+`python examples/render_readme_progression.py --only-sparse` to reproduce them.
+The examples demonstrate mechanics, not measured abuse accuracy.
+
+## Source policies
+
+All new source fields belong to a schema-1.1 dataset. The same canonical request
+works through `analyze`, `analyze_relationships`, CLI JSON and read-only MCP.
+`analyze_series` also accepts a `DatasetV11` or a dictionary containing policy,
+coverage, rules or revisions and returns a schema-1.1 result.
+
+```python
+from anomalyzer import analyze, DataPolicy, DatasetV11
+
+source = DatasetV11(
+    id="calls", entity={"customer": "synthetic"}, units="calls",
+    timestamps=["2026-01-01T00:00:00Z", "2026-01-01T01:00:00Z",
+                "2026-01-01T03:00:00Z"],
+    frequency="1h", values=[2, None, 4],
+    data_policy=DataPolicy(),  # preserve missing; hour 2 is also unknown
+)
+result = analyze({"schema_version": "1.1", "datasets": [source.model_dump(mode="json")]})
+print(result.dataset_results[0].data_quality["samples"])
+```
+
+| Input state | Meaning and behavior |
+| --- | --- |
+| Numeric value, including zero, `coverage="complete"` | A measurement eligible for rules and model support. Explicit zero remains zero. |
+| Null or an omitted internal grid position | Unknown by default; not a zero, observation, calibration residual or recovery. |
+| `coverage="empty"` | Caller confirms a finalized period with no reported volume. Preserve as missing unless an additive zero policy explicitly authorizes conversion. |
+| `coverage="unknown"` | Collection/exposure is unknown; never filled even under a global empty-volume contract. |
+| `coverage="incomplete"` | Original partial value remains visible but is excluded from completed measurements, references, calibration and totals. |
+| `coverage="excluded"` | A measured value excluded from model fitting/calibration. Its explicit range rule can still be assessed; change comparisons break at it. |
+| `coverage="revised"` | Supplied corrected measurement; preserve its revision and availability provenance. |
+
+`coverage`, `available_at` and legacy `period_statuses` align with the original
+input arrays before chronological sorting. Contradictory incomplete annotations
+are rejected. Unknown values do not become observed zero merely because their
+unit is named `calls` or their numeric neighbor is zero.
+
+### Explicitly justified empty volume
+
+```python
+source = source.model_copy(update={
+    "coverage": ["complete", "empty", "complete"],
+    "data_policy": DataPolicy(measurement="additive_volume", missing="zero_for_empty"),
+})
+# Hour 1 is a confirmed empty zero. Omitted hour 2 remains unknown.
+```
+
+Use `absence_means_zero=True` only when a trusted collection/exposure contract
+confirms that **every otherwise unannotated period inside the supplied grid** is
+complete and absence means no volume. Explicit unknown/incomplete annotations
+override that contract. This option requires both `measurement="additive_volume"`
+and `missing="zero_for_empty"`. It does not infer heartbeats or successful collection.
+For per-call cost, output/input or cache share, a no-call period is undefined;
+the zero-volume policy is inappropriate.
+
+Preparation reports original values, model values, coverage, source indexes,
+conversion reasons, counts, revision IDs and availability timestamps. Converted
+zeros are visibly identified as policy transformations. No expectation,
+interpolation or carried-forward value trains the model as a measurement.
+For policy-prepared sources, `grid_count` counts positions, `observation_count`
+counts usable complete values (including authorized zero volume and measured
+model exclusions), `reported_measurement_count` counts actually reported numeric
+measurements, and `usable_model_count` excludes model-masked values. Training and
+calibration expose their own eligible indexes/counts. Grid size is not training support.
+
+## Calendar coordinates and bounded grids
+
+The declared frequency is a fixed elapsed cadence. Internal omitted positions
+are expanded on that grid and handled exactly like nulls. `grid_start` and
+`grid_end` optionally declare its bounds, including leading/trailing empty periods;
+otherwise bounds come from the supplied as-of-available timestamp range.
+Nothing outside those bounds is invented. Off-grid values, off-grid bounds and
+contradictory inputs fail visibly. No cadence is guessed.
+
+The four-dataset/four-relationship limits remain. Raw samples plus revisions and
+the expanded source grids share `max_points` (default 10,000). The grid size is
+checked before allocation, and preparation/model loops use the runtime budget.
+Partition large histories into compatible bounded requests without moving the
+training origin. A failed or unsupported sibling does not erase other results.
+Full evidence can be several megabytes for a bounded two-source history; the
+point limit is not a guarantee of a small result. The standalone
+`examples/library/benchmark_sparse.py` records warm runtime, serialization time
+and result bytes for 1,344-hour and bounded event workloads. CI runs it in the
+offline Lambda-compatible image and retains the report beside its exact image
+and native dependency wheel hashes. Measure Guard's actual memory, concurrency,
+cold start and persistence costs; store a suitable compact application evidence
+contract rather than repeatedly transporting every raw result on each refresh.
+
+Duplicate timestamps normally fail. Legacy schema-1.1 relationship preparation
+retains its unique-source requirement. With an explicit `data_policy`, complete,
+unrevisioned calendar duplicates can follow the declared `duplicate_policy`
+(`sum` or `mean`); original values/indexes remain in `duplicate_sources`. Nulls
+propagate. Duplicates involving incomplete coverage or revision metadata are
+rejected. Event-index timestamps must be unique; ties are refused rather than
+assigned invented times or counted as multiple independent events.
+
+## What sparse statistical support means
+
+Schema-1.1 gaps select an inspectable sparse profile. An explicit `data_policy`
+also selects it for dense data. The default `seasonal-residual-v1` sparse profile
+collects `training_size` **eligible observations**, retaining their actual grid
+indexes. For each declared phase it freezes the training median (mean with
+`outlier_handling="include"`). A phase needs `minimum_phase_samples`, default 2.
+It then collects `calibration_size` eligible forecast residuals, not elapsed slots.
+Unknown/incomplete/excluded positions never count toward either requirement.
+Robust calibration screening is reported separately. Only subsequent calibrated
+points can cross statistical thresholds; early/provisional evidence cannot trigger.
+
+References match `(grid index - regime origin) % season_length`, so missing
+Tuesday does not turn Wednesday into Tuesday. Training origins, indexes, phase
+support, calibration indexes/cutoff and exclusions are inspectable. Unsupported
+phases stay unavailable with a reason. Trend `auto` resolves visibly to no trend
+in this frozen sparse profile. Evaluation samples never update that reference,
+so gap handling cannot learn away a suspected sustained episode.
+
+`reference_gaps="same_phase"` permits a point reference to borrow supported earlier
+training phases across a gap. `reference_gaps="restart"` restarts training and
+calibration after gaps. Explicit `reset_points` always start new regimes.
+Nelson/CUSUM/moving-range runs and point episodes cannot bridge missing,
+incomplete, excluded or reset boundaries, regardless of point-reference policy.
+Location-pattern assessments and cases remain numerical evidence.
+
+Adaptive recipes and explicit linear/exponential trend fitting currently use
+**complete contiguous spans only**: they restart support at gaps, retaining their
+original calendar indexes and timestamps. Post-gap learning requires an explicit
+reset or caller-selected `reference_gaps="restart"`; the default does not silently
+train a new reference on a suspected episode after unknown collection. Leading
+unavailability before any observed history can precede the initial training span.
+These methods do not borrow references through a gap.
+The existing adaptive policy may admit a persistent third extreme observation;
+this release does not implement an open-episode quarantine policy or simultaneous
+multiple seasonal cycles. Do not label those roadmap features as delivered.
+
+Readiness/reference quality, calibration eligibility, confidence and business
+severity remain different concepts. Sparse readiness is conservatively `caution`
+or `not_assessed`; neither is a probability or a universal switch disabling a
+valid independently measured rule.
+
+## Comparable event-index observations
+
+```python
+from anomalyzer import analyze_series
+
+events = DatasetV11(
+    id="cost_per_call", entity={"customer": "synthetic"}, units="dollars_per_call",
+    timestamps=["2026-01-01T00:00:00Z", "2026-01-03T14:20:00Z"],
+    values=[0.10, 0.12],  # one measurement per genuine comparable call
+    data_policy=DataPolicy(observation_basis="event_index", change_max_gap_seconds=86400),
+)
+result = analyze_series(events, {"season_length": 1, "trend": "none"})
+```
+
+Omit `frequency`: event position is not elapsed time. Original event timestamps
+remain on assessments and lineage. Calendar seasonality (`season_length>1`),
+calendar rates, zero filling and calendar grid bounds are unavailable on this
+basis. A caller-declared maximum time separation can limit change comparisons.
+Without it, successive known comparable events are the explicit comparison basis.
+Unknown event measurements break comparisons. Values not yet available at the
+as-of cutoff do not create known event positions.
+Event patterns refer to recorded comparable events, not uninterrupted calendar
+coverage. Recorded events do not establish complete collection between them;
+declare an explicit reset after an unknown collection boundary rather than
+inventing a call to mark it.
+
+Comparability of tenant, operation, model, execution context and sample unit is
+the caller's responsibility. Tokens, records or consumed units are not implicitly
+independent events. This mode does not infer abuse, invocation count or timing
+resolution that the telemetry never recorded.
+
+## Rules before learning; relationships with gaps
+
+A measured `acceptable_range` rule can be violated on the first usable value,
+even when statistical status is `insufficient_history` or `inapplicable`.
+`maximum_absolute_change` and `maximum_relative_change` require a comparable prior
+value. Unknown/incomplete/excluded positions and explicit resets break that prior.
+Relative change from zero is unavailable. An optional `change_max_gap_seconds`
+also limits separation. Unavailable rule reasons are included in assessment
+baselines. A descriptive candidate alone remains non-triggering.
+
+Relationships align valid original coordinates and may use different source
+missing-data policies. A missing input is unavailable; a zero denominator remains
+undefined; nonfinite derived arithmetic is unavailable rather than invented.
+Lineage records source values, coverage, indexes, revision IDs and availability.
+An excluded but genuinely measured value can still support exact arithmetic and
+range assessment, while its derived model sample stays excluded.
+
+## Sparse multi-resolution views
+
+Schema 1.1 `multi-resolution-v1` exposes distinct `intraday` and `completed_period`
+targets, assessments, cases and evidence IDs. The child recipe is explicitly
+frozen seasonal analysis. Native evidence remains available when a daily view
+lacks support. Daily sources require declared additive volume and `sum`; all
+expected fixed-UTC-day children must be valid and complete. A missing or
+open child makes its completed total unavailable. A reset inside a fully measured
+day or an excluded model child excludes that day's model sample; the real sum
+can still support a separately declared daily range rule. Observed
+partial sums are progress metadata, never a completed total. Empty days are not
+silently compressed out of the daily grid.
+
+Daily ratios divide summed compatible numerator and denominator quantities;
+percentages are not averaged. Daily differences use compatible additive sums.
+Daily lagged, normalized-residual and joint-condition definitions remain
+explicitly inapplicable until their per-view settings are supported. Hourly
+lags/scales/thresholds are not silently reinterpreted as daily ones.
+
+Source and relationship `rules` apply natively. Use `completed_period_rules` for
+separately meaningful daily rules. View cases remain separate on policy replay;
+they are correlated evidence from shared source samples, not independent votes.
+Local-calendar/DST day aggregation and adaptive multi-resolution children remain
+unsupported. Schema-1.0 multi-resolution retains its earlier complete-data contract.
+
+## Availability, revisions and recomputation
+
+`available_at` optionally records when each original value became known.
+`revisions` contains explicit `{timestamp, value, available_at, revision_id, coverage}`
+records. Revisions need distinct event-time/availability pairs. As-of analysis
+selects only values/revisions available by the cutoff, and the latest eligible
+revision wins. A revised snapshot with no earlier version cannot reconstruct the
+missing historical value. Revision coverage is an input fact, not persistent
+state maintained by the library.
+
+Case availability cutoffs conservatively retain supplied known-input timing.
+Unknown arrival times remain event-time snapshots and cannot establish
+delivery-time detection latency. Later corrections can change a later recomputed
+result; they must not rewrite an earlier as-of assessment. Keep the same grid
+origin, training/calibration history, profile revision and reviewed reset boundaries
+when recomputing a bounded history. Moving the request start can relearn a different
+reference. Durable append/state storage and automatic correction reconciliation
+remain caller responsibilities and separate roadmap work.
+
+## Guard adoption
+
+Install the exact [immutable release and hash-pinned dependencies](releases/README.md)
+at application build time in Linux/Python 3.12 for the Lambda's actual architecture.
+Bundle Pydantic's compatible native wheel and all other dependencies. Execution
+requires no checkout, CLI, MCP process, hosted engine or package download.
+
+Guard owns collection/exposure declarations, metric units, trusted attribution,
+profile selection, fixed history origins, persistence, scheduling, business
+materiality and Warning/Critical severity, deduplication/recovery and opt-in
+notifications. API/Agentic Calls, LLM dollars and custom-meter units remain distinct.
+No-data and failed checks cannot recover an incident. This library release does
+not integrate or deploy Guard, invent risk thresholds, authorize blocking, or send
+notifications. Hourly history cannot reconstruct minute activity.
diff --git a/examples/README.md b/examples/README.md
index 0804055..5d3b220 100644
--- a/examples/README.md
+++ b/examples/README.md
@@ -1,5 +1,12 @@
 # Try the anomaly detector
 
+The README progression continues through chapters 14–22 with
+[sparse-data fixtures](sparse/), explicit empty policies, coverage, event-index
+measurements, calendar phase, gap-broken patterns, early rules and incomplete-day
+relationships. `generate.py` generates these fixtures, `run_demo.py` checks them,
+and `render_readme_progression.py` renders their figures in the established style.
+See the [source-policy API guide](../docs/sparse-data.md) for interpretation limits.
+
 From the project root:
 
 ```powershell
diff --git a/examples/generate.py b/examples/generate.py
index d8278c9..40fcc69 100644
--- a/examples/generate.py
+++ b/examples/generate.py
@@ -1,4 +1,5 @@
 """Generate deterministic, synthetic demonstrations; no production effectiveness claim."""
+from sparse_examples import generate as generate_sparse
 import csv
 import json
 import math
@@ -6,6 +7,7 @@
 from pathlib import Path
 
 base = Path(__file__).parent
+generate_sparse()
 ordinary = [100 + [0, 10, -5, 8, 2, -12, -8][i % 7] + [0.2, -0.4, 0.6, -0.3, 0.1][i % 5] for i in range(90)]
 for name in ("ordinary", "spike", "drop", "shift", "growth_up", "growth_down",
              "compound", "compound_down", "location_shift_up", "location_shift_down",
@@ -50,7 +52,11 @@
                      150 - 0.4*(i-50) if i < 100 else
                      105 * 1.01**(i-100))
             values.append(round(level + weekly[i % 7] + [0.2, -0.4, 0.6, -0.3, 0.1][i % 5], 4))
-    config = {"season_length": 7, "max_runtime_seconds": 120}
+    config = {"season_length": 7}
+    if name in ("shift", "steps"):
+        # Preserve the original progression's single-reference sustained-shift profile.
+        config["robust_reference_seasons"] = 1
+    config["max_runtime_seconds"] = 120
     if name.startswith("compound"):
         # Five complete seasonal cycles make the synthetic exponential rate
         # identifiable despite the small deterministic variation.
diff --git a/examples/library/benchmark_sparse.py b/examples/library/benchmark_sparse.py
new file mode 100644
index 0000000..2d241bf
--- /dev/null
+++ b/examples/library/benchmark_sparse.py
@@ -0,0 +1,75 @@
+"""Bounded synthetic consumer benchmark; runs against an installed offline package."""
+import argparse
+from copy import deepcopy
+from datetime import datetime, timedelta, timezone
+import json
+from pathlib import Path
+import platform
+import statistics
+import time
+
+from anomalyzer import analyze, __version__
+
+
+def workloads():
+    count = 1344
+    times = [(datetime(2026,1,1,tzinfo=timezone.utc)+timedelta(hours=i)).isoformat() for i in range(count)]
+    values = [1.0+(i%3)*.1 if i%13 == 0 else None for i in range(count)]
+    ds = {"id":"volume", "entity":{"customer":"synthetic"}, "units":"calls", "timestamps":times,
+          "frequency":"1h", "values":values, "data_policy":{}}
+    base = {"schema_version":"1.1", "datasets":[ds],
+            "config":{"season_length":1,"training_size":28,"calibration_size":14,"trend":"none"}}
+    result = {"1344h_preserve_unknown":deepcopy(base)}
+    zero = deepcopy(base)
+    zero["datasets"][0]["data_policy"] = {"measurement":"additive_volume","missing":"zero_for_empty","absence_means_zero":True}
+    result["1344h_confirmed_empty_volume"] = zero
+    relationship = deepcopy(zero)
+    relationship["datasets"][0]["id"] = "cost"
+    relationship["datasets"][0]["units"] = "dollars"
+    relationship["datasets"].append({**deepcopy(zero["datasets"][0]), "id":"calls", "values":[1 if v is not None else None for v in values]})
+    relationship["relationships"] = [{"id":"unit-cost","kind":"ratio","numerator":"cost","denominator":"calls","units":"dollars_per_call"}]
+    result["1344h_two_sources_and_ratio"] = relationship
+    multi = deepcopy(relationship)
+    multi["config"].update(recipe="multi-resolution-v1",completed_period_season_length=7)
+    result["1344h_two_source_multi_resolution"] = multi
+    events = deepcopy(base)
+    events["datasets"][0].update(timestamps=[times[i] for i in range(0,count,3)],
+        values=[.10+(i%5)*.001 for i in range(0,count,3)], units="dollars_per_call",
+        data_policy={"observation_basis":"event_index"})
+    events["datasets"][0].pop("frequency")
+    result["448_comparable_events"] = events
+    return result
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--output", type=Path)
+    parser.add_argument("--repeats", type=int, default=3)
+    args = parser.parse_args()
+    assert 1 <= args.repeats <= 10
+    rows = []
+    for name, request in workloads().items():
+        analyze(request)
+        samples = []
+        for _ in range(args.repeats):
+            started = time.perf_counter()
+            result = analyze(request)
+            samples.append(time.perf_counter()-started)
+        started = time.perf_counter()
+        encoded = result.model_dump_json().encode()
+        serialization = time.perf_counter()-started
+        assert not any(item.status == "failed" for item in result.dataset_results), name
+        rows.append({"workload":name,"input_positions":sum(len(d["values"]) for d in request["datasets"]),
+            "median_ms":round(statistics.median(samples)*1000,2), "serialization_ms":round(serialization*1000,2),
+            "result_bytes":len(encoded), "status":result.status})
+    report = {"version":__version__,"environment":platform.platform(),"python":platform.python_version(),
+        "repeats":args.repeats,"rows":rows,
+        "limits":"Synthetic warm process timings exclude imports, scheduling, cold starts and persistence. No accuracy, capacity or production cost claim."}
+    if args.output:
+        args.output.parent.mkdir(parents=True, exist_ok=True)
+        args.output.write_text(json.dumps(report,indent=2)+"\n",encoding="utf-8")
+    print(json.dumps(report,indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/examples/library/guard_smoke.py b/examples/library/guard_smoke.py
index 4348481..debc5c4 100644
--- a/examples/library/guard_smoke.py
+++ b/examples/library/guard_smoke.py
@@ -103,6 +103,21 @@ def offline(*args, **kwargs):
                      "config": {"recipe": "multi-resolution-v1", "training_size": 2,
                                 "calibration_size": 3, "completed_period_season_length": 1}})
     assert multi.data_quality["incomplete_period"]["included_in_completed_period_view"] is False
+    # New source contracts must work from the distribution, without local fixtures.
+    sparse = {"schema_version":"1.1", "datasets":[{
+        "id":"calls","entity":{"customer":"synthetic"},"units":"calls",
+        "timestamps":timestamps[:12],"frequency":"1d","values":[1,None,0,1,1,0,1,1,None,1,0,20],
+        "coverage":["complete","empty"]+["complete"]*6+["unknown"]+["complete"]*3,
+        "data_policy":{"measurement":"additive_volume","missing":"zero_for_empty"},
+        "rules":[{"id":"measured-limit","kind":"acceptable_range","minimum":0,"maximum":10}]}],
+        "config":{"season_length":1,"training_size":4,"calibration_size":3,"trend":"none"}}
+    result_sparse = analyze_relationships(sparse)
+    quality = result_sparse.dataset_results[0].data_quality
+    assert quality["zero_filled_count"] == 1 and quality["samples"][8]["value"] is None
+    assert result_sparse.cases
+    events = {**sparse["datasets"][0], "data_policy":{"observation_basis":"event_index"}}
+    events.pop("frequency")
+    assert analyze_series(events, sparse["config"]).dataset_results[0].data_quality["coordinate"] == "event_index"
     print(json.dumps({"version": anomalyzer.__version__, "status": result.status,
                       "cases": len(result.cases), "offline": True,
                       "pydantic_core": pydantic_core.__file__}))
diff --git a/examples/render_readme_progression.py b/examples/render_readme_progression.py
index e69a2df..1a0eebf 100644
--- a/examples/render_readme_progression.py
+++ b/examples/render_readme_progression.py
@@ -161,7 +161,10 @@ def chart(svg, values, overlays, x, y, width, height, *, y_bounds=None,
         points = [None if value is None else (sx(index), sy(value))
                   for index, value in enumerate(series)]
         svg.polyline(points, color, line_width, dash, opacity)
-    svg.polyline([(sx(index), sy(value)) for index, value in enumerate(values)], BLUE, 2.5)
+    svg.polyline([None if value is None else (sx(index), sy(value)) for index, value in enumerate(values)], BLUE, 2.5)
+    for index, value in enumerate(values):
+        if value is not None and (index == 0 or values[index-1] is None) and (index == count-1 or values[index+1] is None):
+            svg.circle(sx(index), sy(value), 3, BLUE)
 
     for index in flags:
         color = RED if index in injected else ORANGE if index in echoes else RED
@@ -841,6 +844,153 @@ def find_browser():
     raise RuntimeError("No Chromium browser found; set ANOMALYZER_BROWSER to Chrome, Chromium, or Edge")
 
 
+def sparse_progression():
+    from sparse_examples import fixtures
+    requests = fixtures()
+    results = {name: analyze(value) for name, value in requests.items()}
+    figures = {}
+    def target(name):
+        return results[name].dataset_results[0]
+    def observed(name):
+        return [row["original_value"] if row["coverage"] in ("complete", "revised") else None
+                for row in target(name).data_quality["samples"]]
+    def expectation(name):
+        item = target(name)
+        count = item.data_quality["grid_count"]
+        refs = {row["index"]: row["expected"] for m in item.methods for row in m.evidence}
+        refs.update({row["index"]: row["modeled_expectation"] for m in item.methods
+                     for row in m.diagnostics.get("unavailable", []) if "modeled_expectation" in row})
+        return [refs.get(i) for i in range(count)]
+    def panel(svg, name, y, label, height=145, y_bounds=None):
+        values = observed(name)
+        modeled = expectation(name)
+        numbers = [v for v in values + modeled if v is not None]
+        if target(name).data_quality["zero_filled_count"]:
+            numbers.append(0)
+        low, high = min(numbers), max(numbers)
+        pad = max((high-low)*.12, max(abs(low), abs(high))*.02, .001)
+        limits = y_bounds or (max(0, low-pad) if low >= 0 else low-pad, high+pad)
+        flags = {row["index"] for m in target(name).methods for row in m.evidence if row["triggers"] and values[row["index"]] is not None}
+        axes = chart(svg, values, [(NAVY, modeled, 2, "7 5", 1)],
+            95, y, 1035, height, shade_until=None, flags=flags, y_bounds=limits,
+            y_tick_format=(lambda value: f"{value:.2f}") if high < 2 else None,
+            ticks=[(0, "position 0"), (len(values)//2, f"position {len(values)//2}"), (len(values)-1, f"position {len(values)-1}")], label=label)
+        return axes
+    def footer(svg, text, zeros=False):
+        legend(svg, [("line", BLUE, "measured"), ("dash", NAVY, "modeled expectation"), ("dot", RED, "point criterion")] +
+               ([("dot", GOLD, "policy-derived zero")] if zeros else []), y=625)
+        svg.text(60, 658, text, 14, MUTED)
+    svg = Svg()
+    header(svg, "14", "Choose what an empty period means", "Identical source timestamps and values; only the declared empty-data policy changes.")
+    for name, y, title in (("preserve", 175, "Default: confirmed empty remains missing"),
+                           ("zero_empty", 410, "Additive volume + explicit zero-for-empty policy")):
+        shared_high = max(v for v in observed("preserve")+expectation("preserve")+observed("zero_empty")+expectation("zero_empty") if v is not None)*1.12
+        sx, sy = panel(svg, name, y, title, y_bounds=(0, shared_high))
+        if name == "zero_empty":
+            for row in target(name).data_quality["samples"]:
+                if row["reason"] == "confirmed_empty_additive_volume":
+                    svg.circle(sx(row["index"]), sy(0), 5, "#ffffff", GOLD, 2)
+        svg.text(1120, y-17, f"{target(name).data_quality['zero_filled_count']} conversions", 15, MUTED, anchor="end")
+    footer(svg, "A collection/exposure declaration makes an empty count zero. Unknown collection never supplies that declaration.", zeros=True)
+    figures["anomaly-progression-14-empty-policy"] = svg.finish()
+    svg = Svg()
+    header(svg, "15", "Keep zero, unknown and unfinished distinct", "Blue zeros are actual measurements; gaps stay gaps; the orange open-period value is excluded.")
+    sx, sy = panel(svg, "statuses", 190, "Calls on the original hourly grid", 320)
+    for i in (1, 7):
+        svg.circle(sx(i), sy(0), 5, BLUE)
+    for i in (2, 9):
+        svg.line(sx(i), 185, sx(i), 515, AXIS, 2, "3 6")
+        svg.text(sx(i), 560, "unknown", 15, MUTED, anchor="middle")
+    svg.circle(sx(13), sy(3), 6, "#ffffff", ORANGE, 2)
+    svg.text(1120, sy(3)-18, "3 so far; open", 16, ORANGE, anchor="end")
+    footer(svg, "No value or finding in a missing period establishes normal behavior or recovery.")
+    figures["anomaly-progression-15-coverage-states"] = svg.finish()
+    svg = Svg()
+    header(svg, "16", "Retain a quiet customer's later event", "The collection contract confirms quiet volume periods; the 12-hour reference keeps their actual phase.")
+    sx, sy = panel(svg, "quiet_volume", 185, "Synthetic additive call volume", 345)
+    for row in target("quiet_volume").data_quality["samples"]:
+        if row["reason"] == "confirmed_empty_additive_volume":
+            svg.circle(sx(row["index"]), sy(0), 2.5, GOLD)
+    svg.text(1115, 165, "Later measurement: 40 calls", 18, RED, 700, "end")
+    footer(svg, "Synthetic numerical evidence only; the calling application decides business materiality and severity.", zeros=True)
+    figures["anomaly-progression-16-quiet-volume"] = svg.finish()
+    svg = Svg()
+    header(svg, "17", "Compare real calls across quiet hours", "A per-call value is undefined in a no-call hour. Event position is an explicitly different observation basis.")
+    panel(svg, "cost_calendar", 175, "Calendar grid: 431 hours; daily-phase support unavailable")
+    panel(svg, "cost_events", 390, "Event index: 12 comparable calls; original event timestamps retained")
+    svg.text(95, 585, "Event 0: Jan 1 00:00 UTC  ·  event 11: Jan 18 22:00 UTC  ·  no elapsed-time rate inferred", 16, MUTED)
+    footer(svg, "Event-index profiles require season_length=1. Workload/model compatibility and true event units are caller declarations.")
+    figures["anomaly-progression-17-event-index"] = svg.finish()
+    svg = Svg()
+    header(svg, "18", "Keep the weekday when observations are absent", "Calendar grid slots remain in place; frozen reference phases use only eligible training observations.")
+    sx, _ = panel(svg, "weekday_phase", 190, "Daily calls; actual seven-day phase", 330)
+    for index in (8, 21, 35, 50, 69):
+        svg.line(sx(index), 185, sx(index), 525, AXIS, 1, "3 6")
+    last = target("weekday_phase").methods[0].evidence[-1]
+    svg.text(1120, 165, f"Latest: {last['observed']:.1f}; same-phase reference: {last['expected']:.1f}", 17, RED, 700, "end")
+    footer(svg, "Dashed expectations may cross a gap; the measured series never does. No evaluation point is admitted to this frozen reference.")
+    figures["anomaly-progression-18-sparse-phase"] = svg.finish()
+    svg = Svg()
+    header(svg, "19", "Restart consecutive patterns at a gap", "Four positive residuals, one unknown hour, then seven positive residuals do not make eleven consecutive observations.")
+    sx, _ = panel(svg, "pattern_gap", 190, "Frozen reference; point threshold intentionally high", 310)
+    svg.line(sx(11), 175, sx(11), 530, AXIS, 3, "4 6")
+    svg.text(sx(11), 560, "unknown hour: pattern boundary", 16, MUTED, anchor="middle")
+    hits = [p["detection_index"] for p in target("pattern_gap").anomaly_patterns if p["rule"] == "cusum"]
+    svg.text(95, 595, f"CUSUM restarts independently: detections at {hits}. Nelson Rule 2 (nine consecutive): unavailable here.", 16, INK)
+    footer(svg, "Missing periods and explicit resets break pattern runs; supported earlier same-phase point references can remain usable.")
+    figures["anomaly-progression-19-pattern-gap"] = svg.finish()
+    svg = Svg()
+    header(svg, "20", "Apply a measured rule on the first observation", "The numerical range is caller-supplied; no learned distribution or business-risk threshold is invented.")
+    scale = lambda value: 520-value/110*360
+    svg.rect(110, scale(20), 560, 520-scale(20), "#eef4ef")
+    svg.line(110, scale(20), 670, scale(20), GOLD, 2, "6 5")
+    svg.line(110, 160, 110, 520, AXIS)
+    svg.line(110, 520, 670, 520, AXIS)
+    svg.circle(390, scale(100), 8, RED)
+    svg.text(390, scale(100)-23, "Observed: 100 measured units", 20, RED, 700, "middle")
+    svg.text(130, scale(20)+29, "Declared acceptable range: 0 to 20", 19, INK)
+    for value in (0, 20, 100):
+        svg.text(95, scale(value)+5, value, 15, MUTED, anchor="end")
+    for y, text in ((210, "Statistical learning: insufficient history"), (270, "Explicit range criterion: met"),
+                    (330, "Numerical policy eligibility: enabled"), (390, "Business severity: caller responsibility")):
+        svg.text(720, y, text, 18, INK)
+    svg.text(390, 558, "First usable observation", 18, MUTED, anchor="middle")
+    svg.text(60, 650, "This synthetic threshold illustrates rule mechanics, not abuse accuracy, incident certainty or permission to act.", 16, MUTED)
+    figures["anomaly-progression-20-first-rule"] = svg.finish()
+    svg = Svg()
+    header(svg, "21", "Keep undefined ratios unavailable", "Aligned quantities are required: a missing input or a zero denominator never produces an observed zero ratio.")
+    req = requests["relationships"]
+    ratio = results["relationships"].relationship_results[0]
+    for values, y, title in ((req["datasets"][0]["values"], 175, "Numerator: actual dollars"),
+                              (req["datasets"][1]["values"], 365, "Denominator: actual calls"),
+                              ([r.value for r in ratio.lineage], 535, "Derived cost per call; gaps are undefined")):
+        chart(svg, values, [], 95, y, 1035, 85, shade_until=None,
+              y_bounds=(0, 1) if y == 535 else None,
+              y_tick_format=(lambda value: f"${value:.2f}") if y == 535 else None,
+              ticks=[(0,"hour 0"),(5,"hour 5"),(11,"hour 11")], label=title)
+    svg.text(60, 665, "At hour 1: denominator zero. At hours 2, 9: missing source. Each unavailable point retains its source lineage and reason.", 15, MUTED)
+    figures["anomaly-progression-21-undefined-ratios"] = svg.finish()
+    svg = Svg()
+    header(svg, "22", "Preserve incomplete days and useful hourly evidence", "A missing child is not zero consumption; a partial sum is never presented as a completed total.")
+    item = next(d for d in results["unfinished_daily"].dataset_results if d.resolution == "intraday")
+    values = [s["value"] for s in item.data_quality["samples"]]
+    sx, sy = chart(svg, values, [], 95, 175, 1035, 180, y_bounds=(0, 3.5), shade_until=None,
+        ticks=[(0,"Jan 1"),(24,"Jan 2"),(48,"Jan 3"),(59,"hour 59: open")], label="Native hourly observations")
+    svg.circle(sx(59), sy(3), 6, "#ffffff", ORANGE, 2)
+    svg.line(sx(5), 165, sx(5), 360, AXIS, 2, "3 6")
+    daily = next(d for d in results["unfinished_daily"].dataset_results if d.resolution == "completed_period")
+    for index, record in enumerate(daily.data_quality["aggregation"]):
+        x = 70+index*370
+        svg.rect(x, 425, 330, 155, "#eef4ef" if record["complete"] else SHADE, radius=8)
+        svg.text(x+18, 457, f"Day {index+1}: {'complete' if record['complete'] else 'incomplete'}", 20, INK, 700)
+        svg.text(x+18, 495, f"{record['usable_children']} / 24 finalized children", 18, MUTED)
+        svg.text(x+18, 535, f"Total: {record['aggregate_value'] if record['complete'] else 'unavailable'}", 22, NAVY if record["complete"] else ORANGE, 700)
+    svg.text(70, 625, "Hourly evidence: available. Completed-day learning: insufficient history. Views share the same source and are correlated.", 17, INK)
+    svg.text(70, 660, "Fixed UTC 24-hour windows; local-calendar/DST aggregation and adaptive multi-resolution children remain separate work.", 15, MUTED)
+    figures["anomaly-progression-22-sparse-days"] = svg.finish()
+    return figures
+
+
 def render(name, svg, browser, temp):
     page = temp / f"{name}.html"
     page.write_text(
@@ -883,6 +1033,12 @@ def main():
         "nelson-rules-reference": figure_rule_reference(),
         "residual-detectors-reference": figure_extended_rule_reference(),
     }
+    import sys
+    sparse = sparse_progression()
+    if "--only-sparse" in sys.argv:
+        figures = sparse
+    else:
+        figures.update(sparse)
     with tempfile.TemporaryDirectory(prefix="anomalyzer-readme-") as directory:
         temp = Path(directory)
         for name, svg in figures.items():
diff --git a/examples/run_demo.py b/examples/run_demo.py
index a141526..3681dc3 100644
--- a/examples/run_demo.py
+++ b/examples/run_demo.py
@@ -122,6 +122,8 @@ def main():
                   f"actual: {sorted(actual_patterns)}")
     print("\nRobust mode keeps actual incidents visible while preventing their seasonal echoes.")
     print("These synthetic checks demonstrate behavior, not real-world detection accuracy.")
+    from sparse_examples import run_checks
+    run_checks()
     return 0 if passed else 1
 
 
diff --git a/examples/sparse/cost_calendar.json b/examples/sparse/cost_calendar.json
new file mode 100644
index 0000000..68b0a9b
--- /dev/null
+++ b/examples/sparse/cost_calendar.json
@@ -0,0 +1,48 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "dollars_per_call",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-02T11:00:00+00:00",
+        "2026-01-04T08:00:00+00:00",
+        "2026-01-05T04:00:00+00:00",
+        "2026-01-06T10:00:00+00:00",
+        "2026-01-08T12:00:00+00:00",
+        "2026-01-10T04:00:00+00:00",
+        "2026-01-13T12:00:00+00:00",
+        "2026-01-15T14:00:00+00:00",
+        "2026-01-18T22:00:00+00:00"
+      ],
+      "values": [
+        0.1,
+        0.11,
+        0.09,
+        0.1,
+        0.09,
+        0.1,
+        0.11,
+        0.1,
+        0.09,
+        0.1,
+        0.11,
+        0.8
+      ],
+      "data_policy": {},
+      "frequency": "1h"
+    }
+  ],
+  "config": {
+    "season_length": 24,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/cost_events.json b/examples/sparse/cost_events.json
new file mode 100644
index 0000000..88ef135
--- /dev/null
+++ b/examples/sparse/cost_events.json
@@ -0,0 +1,49 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "dollars_per_call",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-02T11:00:00+00:00",
+        "2026-01-04T08:00:00+00:00",
+        "2026-01-05T04:00:00+00:00",
+        "2026-01-06T10:00:00+00:00",
+        "2026-01-08T12:00:00+00:00",
+        "2026-01-10T04:00:00+00:00",
+        "2026-01-13T12:00:00+00:00",
+        "2026-01-15T14:00:00+00:00",
+        "2026-01-18T22:00:00+00:00"
+      ],
+      "values": [
+        0.1,
+        0.11,
+        0.09,
+        0.1,
+        0.09,
+        0.1,
+        0.11,
+        0.1,
+        0.09,
+        0.1,
+        0.11,
+        0.8
+      ],
+      "data_policy": {
+        "observation_basis": "event_index"
+      }
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/first_rule.json b/examples/sparse/first_rule.json
new file mode 100644
index 0000000..b59730f
--- /dev/null
+++ b/examples/sparse/first_rule.json
@@ -0,0 +1,37 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "measured_units",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00"
+      ],
+      "values": [
+        100
+      ],
+      "frequency": "1h",
+      "data_policy": {},
+      "rules": [
+        {
+          "id": "example-range",
+          "kind": "acceptable_range",
+          "minimum": 0,
+          "maximum": 20
+        }
+      ]
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  },
+  "policy": {
+    "allow_explicit_rules": true
+  }
+}
diff --git a/examples/sparse/pattern_gap.json b/examples/sparse/pattern_gap.json
new file mode 100644
index 0000000..5f60186
--- /dev/null
+++ b/examples/sparse/pattern_gap.json
@@ -0,0 +1,63 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00",
+        "2026-01-01T14:00:00+00:00",
+        "2026-01-01T15:00:00+00:00",
+        "2026-01-01T16:00:00+00:00",
+        "2026-01-01T17:00:00+00:00",
+        "2026-01-01T18:00:00+00:00"
+      ],
+      "values": [
+        100,
+        100,
+        100,
+        100,
+        99,
+        100,
+        101,
+        102,
+        102,
+        102,
+        102,
+        null,
+        102,
+        102,
+        102,
+        102,
+        102,
+        102,
+        102
+      ],
+      "frequency": "1h",
+      "data_policy": {}
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none",
+    "point_threshold": 100
+  }
+}
diff --git a/examples/sparse/preserve.json b/examples/sparse/preserve.json
new file mode 100644
index 0000000..cf22411
--- /dev/null
+++ b/examples/sparse/preserve.json
@@ -0,0 +1,82 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00",
+        "2026-01-01T14:00:00+00:00",
+        "2026-01-01T15:00:00+00:00",
+        "2026-01-01T16:00:00+00:00",
+        "2026-01-01T17:00:00+00:00"
+      ],
+      "values": [
+        10,
+        12,
+        11,
+        9,
+        null,
+        10,
+        null,
+        11,
+        9,
+        null,
+        10,
+        12,
+        null,
+        10,
+        11,
+        9,
+        10,
+        30
+      ],
+      "frequency": "1h",
+      "coverage": [
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete"
+      ],
+      "data_policy": {
+        "measurement": "additive_volume"
+      }
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/quiet_volume.json b/examples/sparse/quiet_volume.json
new file mode 100644
index 0000000..fac52e8
--- /dev/null
+++ b/examples/sparse/quiet_volume.json
@@ -0,0 +1,281 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00",
+        "2026-01-01T14:00:00+00:00",
+        "2026-01-01T15:00:00+00:00",
+        "2026-01-01T16:00:00+00:00",
+        "2026-01-01T17:00:00+00:00",
+        "2026-01-01T18:00:00+00:00",
+        "2026-01-01T19:00:00+00:00",
+        "2026-01-01T20:00:00+00:00",
+        "2026-01-01T21:00:00+00:00",
+        "2026-01-01T22:00:00+00:00",
+        "2026-01-01T23:00:00+00:00",
+        "2026-01-02T00:00:00+00:00",
+        "2026-01-02T01:00:00+00:00",
+        "2026-01-02T02:00:00+00:00",
+        "2026-01-02T03:00:00+00:00",
+        "2026-01-02T04:00:00+00:00",
+        "2026-01-02T05:00:00+00:00",
+        "2026-01-02T06:00:00+00:00",
+        "2026-01-02T07:00:00+00:00",
+        "2026-01-02T08:00:00+00:00",
+        "2026-01-02T09:00:00+00:00",
+        "2026-01-02T10:00:00+00:00",
+        "2026-01-02T11:00:00+00:00",
+        "2026-01-02T12:00:00+00:00",
+        "2026-01-02T13:00:00+00:00",
+        "2026-01-02T14:00:00+00:00",
+        "2026-01-02T15:00:00+00:00",
+        "2026-01-02T16:00:00+00:00",
+        "2026-01-02T17:00:00+00:00",
+        "2026-01-02T18:00:00+00:00",
+        "2026-01-02T19:00:00+00:00",
+        "2026-01-02T20:00:00+00:00",
+        "2026-01-02T21:00:00+00:00",
+        "2026-01-02T22:00:00+00:00",
+        "2026-01-02T23:00:00+00:00",
+        "2026-01-03T00:00:00+00:00",
+        "2026-01-03T01:00:00+00:00",
+        "2026-01-03T02:00:00+00:00",
+        "2026-01-03T03:00:00+00:00",
+        "2026-01-03T04:00:00+00:00",
+        "2026-01-03T05:00:00+00:00",
+        "2026-01-03T06:00:00+00:00",
+        "2026-01-03T07:00:00+00:00",
+        "2026-01-03T08:00:00+00:00",
+        "2026-01-03T09:00:00+00:00",
+        "2026-01-03T10:00:00+00:00",
+        "2026-01-03T11:00:00+00:00",
+        "2026-01-03T12:00:00+00:00",
+        "2026-01-03T13:00:00+00:00",
+        "2026-01-03T14:00:00+00:00",
+        "2026-01-03T15:00:00+00:00",
+        "2026-01-03T16:00:00+00:00",
+        "2026-01-03T17:00:00+00:00",
+        "2026-01-03T18:00:00+00:00",
+        "2026-01-03T19:00:00+00:00",
+        "2026-01-03T20:00:00+00:00",
+        "2026-01-03T21:00:00+00:00",
+        "2026-01-03T22:00:00+00:00",
+        "2026-01-03T23:00:00+00:00",
+        "2026-01-04T00:00:00+00:00",
+        "2026-01-04T01:00:00+00:00",
+        "2026-01-04T02:00:00+00:00",
+        "2026-01-04T03:00:00+00:00",
+        "2026-01-04T04:00:00+00:00",
+        "2026-01-04T05:00:00+00:00",
+        "2026-01-04T06:00:00+00:00",
+        "2026-01-04T07:00:00+00:00",
+        "2026-01-04T08:00:00+00:00",
+        "2026-01-04T09:00:00+00:00",
+        "2026-01-04T10:00:00+00:00",
+        "2026-01-04T11:00:00+00:00"
+      ],
+      "values": [
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        1.0,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        null,
+        40
+      ],
+      "frequency": "1h",
+      "coverage": [
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "empty",
+        "complete"
+      ],
+      "data_policy": {
+        "measurement": "additive_volume",
+        "missing": "zero_for_empty"
+      }
+    }
+  ],
+  "config": {
+    "season_length": 12,
+    "training_size": 36,
+    "calibration_size": 24,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/relationships.json b/examples/sparse/relationships.json
new file mode 100644
index 0000000..5613387
--- /dev/null
+++ b/examples/sparse/relationships.json
@@ -0,0 +1,98 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "dollars",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00"
+      ],
+      "values": [
+        1,
+        2,
+        null,
+        3,
+        0,
+        2,
+        4,
+        5,
+        1,
+        3,
+        2,
+        9
+      ],
+      "frequency": "1h",
+      "data_policy": {
+        "measurement": "additive_volume"
+      }
+    },
+    {
+      "id": "calls",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00"
+      ],
+      "values": [
+        10,
+        0,
+        10,
+        10,
+        0,
+        20,
+        40,
+        50,
+        10,
+        null,
+        20,
+        10
+      ],
+      "frequency": "1h",
+      "data_policy": {
+        "measurement": "additive_volume"
+      }
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  },
+  "relationships": [
+    {
+      "id": "unit-cost",
+      "kind": "ratio",
+      "numerator": "series",
+      "denominator": "calls",
+      "units": "dollars_per_call"
+    }
+  ]
+}
diff --git a/examples/sparse/statuses.json b/examples/sparse/statuses.json
new file mode 100644
index 0000000..125d63f
--- /dev/null
+++ b/examples/sparse/statuses.json
@@ -0,0 +1,68 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00"
+      ],
+      "values": [
+        10,
+        0,
+        null,
+        8,
+        10,
+        9,
+        11,
+        0,
+        12,
+        null,
+        9,
+        10,
+        20,
+        3
+      ],
+      "frequency": "1h",
+      "coverage": [
+        "complete",
+        "complete",
+        "unknown",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "unknown",
+        "complete",
+        "complete",
+        "complete",
+        "incomplete"
+      ],
+      "data_policy": {}
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/unfinished_daily.json b/examples/sparse/unfinished_daily.json
new file mode 100644
index 0000000..a32bb98
--- /dev/null
+++ b/examples/sparse/unfinished_daily.json
@@ -0,0 +1,210 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00",
+        "2026-01-01T14:00:00+00:00",
+        "2026-01-01T15:00:00+00:00",
+        "2026-01-01T16:00:00+00:00",
+        "2026-01-01T17:00:00+00:00",
+        "2026-01-01T18:00:00+00:00",
+        "2026-01-01T19:00:00+00:00",
+        "2026-01-01T20:00:00+00:00",
+        "2026-01-01T21:00:00+00:00",
+        "2026-01-01T22:00:00+00:00",
+        "2026-01-01T23:00:00+00:00",
+        "2026-01-02T00:00:00+00:00",
+        "2026-01-02T01:00:00+00:00",
+        "2026-01-02T02:00:00+00:00",
+        "2026-01-02T03:00:00+00:00",
+        "2026-01-02T04:00:00+00:00",
+        "2026-01-02T05:00:00+00:00",
+        "2026-01-02T06:00:00+00:00",
+        "2026-01-02T07:00:00+00:00",
+        "2026-01-02T08:00:00+00:00",
+        "2026-01-02T09:00:00+00:00",
+        "2026-01-02T10:00:00+00:00",
+        "2026-01-02T11:00:00+00:00",
+        "2026-01-02T12:00:00+00:00",
+        "2026-01-02T13:00:00+00:00",
+        "2026-01-02T14:00:00+00:00",
+        "2026-01-02T15:00:00+00:00",
+        "2026-01-02T16:00:00+00:00",
+        "2026-01-02T17:00:00+00:00",
+        "2026-01-02T18:00:00+00:00",
+        "2026-01-02T19:00:00+00:00",
+        "2026-01-02T20:00:00+00:00",
+        "2026-01-02T21:00:00+00:00",
+        "2026-01-02T22:00:00+00:00",
+        "2026-01-02T23:00:00+00:00",
+        "2026-01-03T00:00:00+00:00",
+        "2026-01-03T01:00:00+00:00",
+        "2026-01-03T02:00:00+00:00",
+        "2026-01-03T03:00:00+00:00",
+        "2026-01-03T04:00:00+00:00",
+        "2026-01-03T05:00:00+00:00",
+        "2026-01-03T06:00:00+00:00",
+        "2026-01-03T07:00:00+00:00",
+        "2026-01-03T08:00:00+00:00",
+        "2026-01-03T09:00:00+00:00",
+        "2026-01-03T10:00:00+00:00",
+        "2026-01-03T11:00:00+00:00"
+      ],
+      "values": [
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        null,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        3
+      ],
+      "frequency": "1h",
+      "data_policy": {
+        "measurement": "additive_volume"
+      },
+      "coverage": [
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "incomplete"
+      ]
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none",
+    "recipe": "multi-resolution-v1",
+    "completed_period_season_length": 1
+  }
+}
diff --git a/examples/sparse/weekday_phase.json b/examples/sparse/weekday_phase.json
new file mode 100644
index 0000000..049c634
--- /dev/null
+++ b/examples/sparse/weekday_phase.json
@@ -0,0 +1,204 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-02T00:00:00+00:00",
+        "2026-01-03T00:00:00+00:00",
+        "2026-01-04T00:00:00+00:00",
+        "2026-01-05T00:00:00+00:00",
+        "2026-01-06T00:00:00+00:00",
+        "2026-01-07T00:00:00+00:00",
+        "2026-01-08T00:00:00+00:00",
+        "2026-01-09T00:00:00+00:00",
+        "2026-01-10T00:00:00+00:00",
+        "2026-01-11T00:00:00+00:00",
+        "2026-01-12T00:00:00+00:00",
+        "2026-01-13T00:00:00+00:00",
+        "2026-01-14T00:00:00+00:00",
+        "2026-01-15T00:00:00+00:00",
+        "2026-01-16T00:00:00+00:00",
+        "2026-01-17T00:00:00+00:00",
+        "2026-01-18T00:00:00+00:00",
+        "2026-01-19T00:00:00+00:00",
+        "2026-01-20T00:00:00+00:00",
+        "2026-01-21T00:00:00+00:00",
+        "2026-01-22T00:00:00+00:00",
+        "2026-01-23T00:00:00+00:00",
+        "2026-01-24T00:00:00+00:00",
+        "2026-01-25T00:00:00+00:00",
+        "2026-01-26T00:00:00+00:00",
+        "2026-01-27T00:00:00+00:00",
+        "2026-01-28T00:00:00+00:00",
+        "2026-01-29T00:00:00+00:00",
+        "2026-01-30T00:00:00+00:00",
+        "2026-01-31T00:00:00+00:00",
+        "2026-02-01T00:00:00+00:00",
+        "2026-02-02T00:00:00+00:00",
+        "2026-02-03T00:00:00+00:00",
+        "2026-02-04T00:00:00+00:00",
+        "2026-02-05T00:00:00+00:00",
+        "2026-02-06T00:00:00+00:00",
+        "2026-02-07T00:00:00+00:00",
+        "2026-02-08T00:00:00+00:00",
+        "2026-02-09T00:00:00+00:00",
+        "2026-02-10T00:00:00+00:00",
+        "2026-02-11T00:00:00+00:00",
+        "2026-02-12T00:00:00+00:00",
+        "2026-02-13T00:00:00+00:00",
+        "2026-02-14T00:00:00+00:00",
+        "2026-02-15T00:00:00+00:00",
+        "2026-02-16T00:00:00+00:00",
+        "2026-02-17T00:00:00+00:00",
+        "2026-02-18T00:00:00+00:00",
+        "2026-02-19T00:00:00+00:00",
+        "2026-02-20T00:00:00+00:00",
+        "2026-02-21T00:00:00+00:00",
+        "2026-02-22T00:00:00+00:00",
+        "2026-02-23T00:00:00+00:00",
+        "2026-02-24T00:00:00+00:00",
+        "2026-02-25T00:00:00+00:00",
+        "2026-02-26T00:00:00+00:00",
+        "2026-02-27T00:00:00+00:00",
+        "2026-02-28T00:00:00+00:00",
+        "2026-03-01T00:00:00+00:00",
+        "2026-03-02T00:00:00+00:00",
+        "2026-03-03T00:00:00+00:00",
+        "2026-03-04T00:00:00+00:00",
+        "2026-03-05T00:00:00+00:00",
+        "2026-03-06T00:00:00+00:00",
+        "2026-03-07T00:00:00+00:00",
+        "2026-03-08T00:00:00+00:00",
+        "2026-03-09T00:00:00+00:00",
+        "2026-03-10T00:00:00+00:00",
+        "2026-03-11T00:00:00+00:00",
+        "2026-03-12T00:00:00+00:00",
+        "2026-03-13T00:00:00+00:00",
+        "2026-03-14T00:00:00+00:00",
+        "2026-03-15T00:00:00+00:00",
+        "2026-03-16T00:00:00+00:00",
+        "2026-03-17T00:00:00+00:00",
+        "2026-03-18T00:00:00+00:00",
+        "2026-03-19T00:00:00+00:00",
+        "2026-03-20T00:00:00+00:00",
+        "2026-03-21T00:00:00+00:00",
+        "2026-03-22T00:00:00+00:00",
+        "2026-03-23T00:00:00+00:00",
+        "2026-03-24T00:00:00+00:00",
+        "2026-03-25T00:00:00+00:00",
+        "2026-03-26T00:00:00+00:00",
+        "2026-03-27T00:00:00+00:00",
+        "2026-03-28T00:00:00+00:00",
+        "2026-03-29T00:00:00+00:00",
+        "2026-03-30T00:00:00+00:00",
+        "2026-03-31T00:00:00+00:00"
+      ],
+      "values": [
+        100.2,
+        109.6,
+        95.6,
+        107.7,
+        102.1,
+        88.2,
+        91.6,
+        100.6,
+        null,
+        95.1,
+        108.2,
+        101.6,
+        88.6,
+        91.7,
+        100.1,
+        110.2,
+        94.6,
+        108.6,
+        101.7,
+        88.1,
+        92.2,
+        null,
+        110.6,
+        94.7,
+        108.1,
+        102.2,
+        87.6,
+        92.6,
+        99.7,
+        110.1,
+        95.2,
+        107.6,
+        102.6,
+        87.7,
+        92.1,
+        null,
+        109.6,
+        95.6,
+        107.7,
+        102.1,
+        88.2,
+        91.6,
+        100.6,
+        109.7,
+        95.1,
+        108.2,
+        101.6,
+        88.6,
+        91.7,
+        100.1,
+        null,
+        94.6,
+        108.6,
+        101.7,
+        88.1,
+        92.2,
+        99.6,
+        110.6,
+        94.7,
+        108.1,
+        102.2,
+        87.6,
+        92.6,
+        99.7,
+        110.1,
+        95.2,
+        107.6,
+        102.6,
+        87.7,
+        null,
+        100.2,
+        109.6,
+        95.6,
+        107.7,
+        102.1,
+        88.2,
+        91.6,
+        100.6,
+        109.7,
+        95.1,
+        108.2,
+        101.6,
+        88.6,
+        91.7,
+        100.1,
+        110.2,
+        94.6,
+        108.6,
+        101.7,
+        123.1
+      ],
+      "frequency": "1d",
+      "data_policy": {}
+    }
+  ],
+  "config": {
+    "season_length": 7,
+    "training_size": 28,
+    "calibration_size": 14,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse/zero_empty.json b/examples/sparse/zero_empty.json
new file mode 100644
index 0000000..2e617ea
--- /dev/null
+++ b/examples/sparse/zero_empty.json
@@ -0,0 +1,83 @@
+{
+  "schema_version": "1.1",
+  "datasets": [
+    {
+      "id": "series",
+      "entity": {
+        "customer": "synthetic"
+      },
+      "units": "calls",
+      "timestamps": [
+        "2026-01-01T00:00:00+00:00",
+        "2026-01-01T01:00:00+00:00",
+        "2026-01-01T02:00:00+00:00",
+        "2026-01-01T03:00:00+00:00",
+        "2026-01-01T04:00:00+00:00",
+        "2026-01-01T05:00:00+00:00",
+        "2026-01-01T06:00:00+00:00",
+        "2026-01-01T07:00:00+00:00",
+        "2026-01-01T08:00:00+00:00",
+        "2026-01-01T09:00:00+00:00",
+        "2026-01-01T10:00:00+00:00",
+        "2026-01-01T11:00:00+00:00",
+        "2026-01-01T12:00:00+00:00",
+        "2026-01-01T13:00:00+00:00",
+        "2026-01-01T14:00:00+00:00",
+        "2026-01-01T15:00:00+00:00",
+        "2026-01-01T16:00:00+00:00",
+        "2026-01-01T17:00:00+00:00"
+      ],
+      "values": [
+        10,
+        12,
+        11,
+        9,
+        null,
+        10,
+        null,
+        11,
+        9,
+        null,
+        10,
+        12,
+        null,
+        10,
+        11,
+        9,
+        10,
+        30
+      ],
+      "frequency": "1h",
+      "coverage": [
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "empty",
+        "complete",
+        "complete",
+        "complete",
+        "complete",
+        "complete"
+      ],
+      "data_policy": {
+        "measurement": "additive_volume",
+        "missing": "zero_for_empty"
+      }
+    }
+  ],
+  "config": {
+    "season_length": 1,
+    "training_size": 4,
+    "calibration_size": 3,
+    "trend": "none"
+  }
+}
diff --git a/examples/sparse_examples.py b/examples/sparse_examples.py
new file mode 100644
index 0000000..b6d1419
--- /dev/null
+++ b/examples/sparse_examples.py
@@ -0,0 +1,104 @@
+"""Reproducible continuation of the README's feature-by-feature progression."""
+from copy import deepcopy
+from datetime import datetime, timedelta, timezone
+import json
+from pathlib import Path
+
+
+def dataset(values, **extra):
+    return {"id": "series", "entity": {"customer": "synthetic"}, "units": "calls",
+            "timestamps": [(datetime(2026, 1, 1, tzinfo=timezone.utc)+timedelta(hours=i)).isoformat() for i in range(len(values))],
+            "values": values, "frequency": "1h", **extra}
+
+
+def request(ds, **config):
+    return {"schema_version": "1.1", "datasets": [ds],
+            "config": {"season_length": 1, "training_size": 4, "calibration_size": 3, "trend": "none", **config}}
+
+
+def fixtures():
+    result = {}
+    values = [10, 12, 11, 9, None, 10, None, 11, 9, None, 10, 12, None, 10, 11, 9, 10, 30]
+    preserve = request(dataset(values, coverage=["empty" if v is None else "complete" for v in values],
+                               data_policy={"measurement": "additive_volume"}))
+    zero = deepcopy(preserve)
+    zero["datasets"][0]["data_policy"]["missing"] = "zero_for_empty"
+    result.update(preserve=preserve, zero_empty=zero)
+    result["statuses"] = request(dataset([10, 0, None, 8, 10, 9, 11, 0, 12, None, 9, 10, 20, 3],
+        coverage=["complete"]*2+["unknown"]+["complete"]*6+["unknown"]+["complete"]*3+["incomplete"], data_policy={}))
+    values = [1.0 if i % 12 == 0 else None for i in range(84)]
+    values[-1] = 40
+    result["quiet_volume"] = request(dataset(values,
+        coverage=["empty" if v is None else "complete" for v in values],
+        data_policy={"measurement": "additive_volume", "missing": "zero_for_empty"}), season_length=12, training_size=36, calibration_size=24)
+    hours = [0, 1, 8, 35, 80, 100, 130, 180, 220, 300, 350, 430]
+    costs = [.10, .11, .09, .10, .09, .10, .11, .10, .09, .10, .11, .80]
+    events = dataset(costs, units="dollars_per_call", data_policy={"observation_basis": "event_index"})
+    events.pop("frequency")
+    events["timestamps"] = [(datetime(2026,1,1,tzinfo=timezone.utc)+timedelta(hours=h)).isoformat() for h in hours]
+    result["cost_events"] = request(events)
+    calendar = deepcopy(events)
+    calendar["frequency"] = "1h"
+    calendar["data_policy"] = {}
+    result["cost_calendar"] = request(calendar, season_length=24)
+    phase = dataset([100+[0,10,-5,8,2,-12,-8][i%7]+[.2,-.4,.6,-.3,.1][i%5] for i in range(90)],
+                    frequency="1d", data_policy={})
+    phase["timestamps"] = [(datetime(2026,1,1,tzinfo=timezone.utc)+timedelta(days=i)).isoformat() for i in range(90)]
+    for index in (8, 21, 35, 50, 69):
+        phase["values"][index] = None
+    phase["values"][-1] += 35
+    result["weekday_phase"] = request(phase, season_length=7, training_size=28, calibration_size=14)
+    result["pattern_gap"] = request(dataset([100]*4+[99,100,101]+[102]*4+[None]+[102]*7, data_policy={}), point_threshold=100)
+    result["first_rule"] = request(dataset([100], data_policy={}, units="measured_units",
+        rules=[{"id": "example-range", "kind": "acceptable_range", "minimum": 0, "maximum": 20}]))
+    result["first_rule"]["policy"] = {"allow_explicit_rules": True}
+    relation = request(dataset([1,2,None,3,0,2,4,5,1,3,2,9], units="dollars", data_policy={"measurement":"additive_volume"}))
+    relation["datasets"].append(dataset([10,0,10,10,0,20,40,50,10,None,20,10], id="calls", data_policy={"measurement":"additive_volume"}))
+    relation["relationships"] = [{"id":"unit-cost","kind":"ratio","numerator":"series","denominator":"calls","units":"dollars_per_call"}]
+    result["relationships"] = relation
+    daily = request(dataset([1.0]*60, data_policy={"measurement":"additive_volume"}), recipe="multi-resolution-v1",
+                    completed_period_season_length=1)
+    daily["datasets"][0]["values"][5] = None
+    daily["datasets"][0]["coverage"] = ["complete"]*59+["incomplete"]
+    daily["datasets"][0]["values"][-1] = 3
+    result["unfinished_daily"] = daily
+    return result
+
+
+def generate():
+    directory = Path(__file__).parent / "sparse"
+    directory.mkdir(exist_ok=True)
+    for name, document in fixtures().items():
+        (directory/f"{name}.json").write_text(json.dumps(document, indent=2)+"\n", encoding="utf-8")
+
+
+def run_checks():
+    from anomalyzer import analyze
+    directory = Path(__file__).parent / "sparse"
+    outputs = {name: analyze(json.loads((directory/f"{name}.json").read_text())) for name in fixtures()}
+    for name, result in outputs.items():
+        assert all(item.status != "failed" for item in result.dataset_results), name
+    preserve, zero = [outputs[n].dataset_results[0] for n in ("preserve", "zero_empty")]
+    assert preserve.data_quality["zero_filled_count"] == 0
+    assert zero.data_quality["zero_filled_count"] == 4
+    assert outputs["statuses"].dataset_results[0].data_quality["samples"][-1]["value"] is None
+    assert outputs["quiet_volume"].cases
+    assert outputs["cost_calendar"].status == "insufficient_history"
+    assert outputs["cost_events"].cases
+    phase = outputs["weekday_phase"].dataset_results[0]
+    last = phase.methods[0].evidence[-1]
+    assert last["index"] == 89 and last["triggers"] == ["point"]
+    assert not any(p["rule"] == "nelson_rule_2" for p in outputs["pattern_gap"].dataset_results[0].anomaly_patterns)
+    assert outputs["first_rule"].status == "insufficient_history" and outputs["first_rule"].cases[0].action_eligible
+    reasons = [row.unavailable_reason for row in outputs["relationships"].relationship_results[0].lineage]
+    assert "missing_source_value" in reasons and "zero_denominator" in reasons
+    daily = next(item for item in outputs["unfinished_daily"].dataset_results if item.resolution == "completed_period")
+    assert [r["complete"] for r in daily.data_quality["aggregation"]] == [False, True, False]
+    for name, result in outputs.items():
+        print(f"PASS sparse/{name}: {result.status}; {len(result.cases)} numerical case(s)")
+    return outputs
+
+
+if __name__ == "__main__":
+    generate()
+    run_checks()
diff --git a/pyproject.toml b/pyproject.toml
index 03e7640..70048d1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -32,6 +32,7 @@ Issues = "https://github.com/PredictabilityAtScale/anomaly-detection/issues"
 
 [project.optional-dependencies]
 dev = ["pytest>=8,<10", "jsonschema>=4,<5"]
+docs = ["markdown-it-py==3.0.0"]
 
 [project.scripts]
 anomalyzer = "anomalyzer.cli:main"
@@ -49,6 +50,6 @@ packages = ["src/anomalyzer"]
 [tool.hatch.build.targets.sdist]
 include = [
     "/src", "/tests", "/schemas", "/examples", "/scripts",
-    "/.agents/skills/anomaly-detection", "/docs/releases",
+    "/.agents/skills/anomaly-detection", "/docs/releases", "/docs/sparse-data.md",
     "/README.md", "/CONTRIBUTING.md", "/LICENSE", "/pyproject.toml", "/uv.lock",
 ]
diff --git a/schemas/request-1.1.schema.json b/schemas/request-1.1.schema.json
index 1ebfdeb..84c11ec 100644
--- a/schemas/request-1.1.schema.json
+++ b/schemas/request-1.1.schema.json
@@ -91,6 +91,99 @@
       "title": "Context",
       "type": "object"
     },
+    "DataPolicy": {
+      "additionalProperties": false,
+      "description": "Explicit source semantics; unknown collection is never implicit zero.",
+      "properties": {
+        "missing": {
+          "default": "preserve",
+          "enum": [
+            "preserve",
+            "zero_for_empty"
+          ],
+          "title": "Missing",
+          "type": "string"
+        },
+        "measurement": {
+          "default": "measurement",
+          "enum": [
+            "measurement",
+            "additive_volume"
+          ],
+          "title": "Measurement",
+          "type": "string"
+        },
+        "absence_means_zero": {
+          "default": false,
+          "title": "Absence Means Zero",
+          "type": "boolean"
+        },
+        "observation_basis": {
+          "default": "calendar",
+          "enum": [
+            "calendar",
+            "event_index"
+          ],
+          "title": "Observation Basis",
+          "type": "string"
+        },
+        "reference_gaps": {
+          "default": "same_phase",
+          "enum": [
+            "same_phase",
+            "restart"
+          ],
+          "title": "Reference Gaps",
+          "type": "string"
+        },
+        "grid_start": {
+          "anyOf": [
+            {
+              "type": "string"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Grid Start"
+        },
+        "grid_end": {
+          "anyOf": [
+            {
+              "type": "string"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Grid End"
+        },
+        "change_max_gap_seconds": {
+          "anyOf": [
+            {
+              "exclusiveMinimum": 0,
+              "type": "number"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Change Max Gap Seconds"
+        },
+        "minimum_phase_samples": {
+          "default": 2,
+          "maximum": 10000,
+          "minimum": 1,
+          "title": "Minimum Phase Samples",
+          "type": "integer"
+        }
+      },
+      "title": "DataPolicy",
+      "type": "object"
+    },
     "DatasetV11": {
       "additionalProperties": false,
       "properties": {
@@ -218,6 +311,94 @@
           },
           "title": "Rules",
           "type": "array"
+        },
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
+        "data_policy": {
+          "anyOf": [
+            {
+              "$ref": "#/$defs/DataPolicy"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null
+        },
+        "coverage": {
+          "anyOf": [
+            {
+              "items": {
+                "enum": [
+                  "complete",
+                  "empty",
+                  "unknown",
+                  "incomplete",
+                  "excluded",
+                  "revised"
+                ],
+                "type": "string"
+              },
+              "type": "array"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Coverage"
+        },
+        "available_at": {
+          "anyOf": [
+            {
+              "items": {
+                "anyOf": [
+                  {
+                    "type": "string"
+                  },
+                  {
+                    "type": "null"
+                  }
+                ]
+              },
+              "type": "array"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Available At"
+        },
+        "revisions": {
+          "items": {
+            "$ref": "#/$defs/SampleRevision"
+          },
+          "title": "Revisions",
+          "type": "array"
         }
       },
       "required": [
@@ -230,6 +411,31 @@
     "DifferenceRelationship": {
       "additionalProperties": false,
       "properties": {
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
         "id": {
           "minLength": 1,
           "title": "Id",
@@ -300,6 +506,31 @@
     "JointConditionRelationship": {
       "additionalProperties": false,
       "properties": {
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
         "id": {
           "minLength": 1,
           "title": "Id",
@@ -409,6 +640,31 @@
     "LaggedResponseRelationship": {
       "additionalProperties": false,
       "properties": {
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
         "id": {
           "minLength": 1,
           "title": "Id",
@@ -564,6 +820,31 @@
     "NormalizedResidualRelationship": {
       "additionalProperties": false,
       "properties": {
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
         "id": {
           "minLength": 1,
           "title": "Id",
@@ -640,6 +921,31 @@
     "RatioRelationship": {
       "additionalProperties": false,
       "properties": {
+        "completed_period_rules": {
+          "items": {
+            "discriminator": {
+              "mapping": {
+                "acceptable_range": "#/$defs/AcceptableRangeRule",
+                "maximum_absolute_change": "#/$defs/MaximumAbsoluteChangeRule",
+                "maximum_relative_change": "#/$defs/MaximumRelativeChangeRule"
+              },
+              "propertyName": "kind"
+            },
+            "oneOf": [
+              {
+                "$ref": "#/$defs/AcceptableRangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumAbsoluteChangeRule"
+              },
+              {
+                "$ref": "#/$defs/MaximumRelativeChangeRule"
+              }
+            ]
+          },
+          "title": "Completed Period Rules",
+          "type": "array"
+        },
         "id": {
           "minLength": 1,
           "title": "Id",
@@ -716,6 +1022,56 @@
       "title": "RatioRelationship",
       "type": "object"
     },
+    "SampleRevision": {
+      "additionalProperties": false,
+      "properties": {
+        "timestamp": {
+          "title": "Timestamp",
+          "type": "string"
+        },
+        "value": {
+          "anyOf": [
+            {
+              "type": "number"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "title": "Value"
+        },
+        "available_at": {
+          "title": "Available At",
+          "type": "string"
+        },
+        "revision_id": {
+          "minLength": 1,
+          "title": "Revision Id",
+          "type": "string"
+        },
+        "coverage": {
+          "default": "revised",
+          "enum": [
+            "complete",
+            "empty",
+            "unknown",
+            "incomplete",
+            "excluded",
+            "revised"
+          ],
+          "title": "Coverage",
+          "type": "string"
+        }
+      },
+      "required": [
+        "timestamp",
+        "value",
+        "available_at",
+        "revision_id"
+      ],
+      "title": "SampleRevision",
+      "type": "object"
+    },
     "Settings": {
       "additionalProperties": false,
       "properties": {
diff --git a/schemas/result-1.1.schema.json b/schemas/result-1.1.schema.json
index 97af4e2..3e6b3fd 100644
--- a/schemas/result-1.1.schema.json
+++ b/schemas/result-1.1.schema.json
@@ -322,6 +322,18 @@
           "title": "Id",
           "type": "string"
         },
+        "resolution": {
+          "anyOf": [
+            {
+              "type": "string"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Resolution"
+        },
         "revision_id": {
           "title": "Revision Id",
           "type": "string"
@@ -509,6 +521,18 @@
           "title": "Dataset Id",
           "type": "string"
         },
+        "resolution": {
+          "anyOf": [
+            {
+              "type": "string"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Resolution"
+        },
         "status": {
           "enum": [
             "completed",
@@ -815,6 +839,65 @@
         "latest_source_cutoff": {
           "title": "Latest Source Cutoff",
           "type": "string"
+        },
+        "source_coverage": {
+          "additionalProperties": {
+            "type": "string"
+          },
+          "title": "Source Coverage",
+          "type": "object"
+        },
+        "source_values": {
+          "additionalProperties": {
+            "anyOf": [
+              {
+                "type": "number"
+              },
+              {
+                "type": "null"
+              }
+            ]
+          },
+          "title": "Source Values",
+          "type": "object"
+        },
+        "source_revision_ids": {
+          "additionalProperties": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ]
+          },
+          "title": "Source Revision Ids",
+          "type": "object"
+        },
+        "source_available_at": {
+          "additionalProperties": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ]
+          },
+          "title": "Source Available At",
+          "type": "object"
+        },
+        "source_input_indexes": {
+          "additionalProperties": {
+            "items": {
+              "type": "integer"
+            },
+            "type": "array"
+          },
+          "title": "Source Input Indexes",
+          "type": "object"
         }
       },
       "required": [
@@ -1007,6 +1090,18 @@
           "title": "Relationship Id",
           "type": "string"
         },
+        "resolution": {
+          "anyOf": [
+            {
+              "type": "string"
+            },
+            {
+              "type": "null"
+            }
+          ],
+          "default": null,
+          "title": "Resolution"
+        },
         "kind": {
           "enum": [
             "ratio",
diff --git a/scripts/export_examples_site.py b/scripts/export_examples_site.py
new file mode 100644
index 0000000..e4b10ba
--- /dev/null
+++ b/scripts/export_examples_site.py
@@ -0,0 +1,64 @@
+"""Publish the existing numbered README progression with its checked-in figures."""
+import argparse
+import html
+from pathlib import Path
+import re
+import shutil
+
+from markdown_it import MarkdownIt
+
+ROOT = Path(__file__).resolve().parents[1]
+SOURCE = "https://github.com/PredictabilityAtScale/anomaly-detection/blob/main/"
+
+
+def page(markdown, destination, title):
+    def target(match):
+        label, link = match.groups()
+        if link.startswith("docs/images/") or link.startswith("images/"):
+            link = "assets/"+Path(link).name
+        elif link == "docs/sparse-data.md":
+            link = "sparse-data.html"
+        elif not link.startswith(("https://", "http://", "#")):
+            link = SOURCE+link.removeprefix("../")
+        return f"[{label}]({link})"
+    markdown = re.sub(r"\[([^\]]*)\]\(([^)]+)\)", target, markdown)
+    renderer = MarkdownIt("default", {"html": False})
+    # Stable heading anchors preserve links in the numbered progression.
+    def heading_open(tokens, index, options, env):
+        text = tokens[index+1].content.lower()
+        anchor = re.sub(r"[^\w -]", "", text).replace(" ", "-")
+        return f'<{tokens[index].tag} id="{html.escape(anchor)}">'
+    renderer.add_render_rule("heading_open", lambda self, tokens, idx, opts, env: heading_open(tokens, idx, opts, env))
+    content = renderer.render(markdown)
+    destination.write_text(f'''<!doctype html><html lang="en"><meta charset="utf-8">
+<meta name="viewport" content="width=device-width,initial-scale=1"><title>{html.escape(title)}</title>
+<style>body{{font:18px/1.6 Arial,sans-serif;color:#172330;max-width:1140px;margin:36px auto;padding:0 24px}}
+a{{color:#2563a6}}nav{{border-bottom:1px solid #dce1e7;padding-bottom:18px}}img{{width:100%;height:auto}}
+pre{{background:#eef1f4;padding:18px;overflow:auto}}code{{font-size:.88em}}table{{border-collapse:collapse;width:100%}}
+th,td{{text-align:left;border:1px solid #dce1e7;padding:10px}}h2,h3{{line-height:1.25;margin-top:2em}}</style>
+<nav><a href="../">Capability maps</a> · <a href="index.html">Numbered examples</a> ·
+<a href="sparse-data.html">Sparse-data API</a> · <a href="{SOURCE}docs/releases/README.md">Release guide</a></nav>
+<main>{content}</main><footer>Synthetic examples establish mechanics, not measured abuse accuracy.</footer></html>''', encoding="utf-8")
+    for image in re.findall(r'src="(assets/[^\"]+)"', content):
+        if not (destination.parent/image).is_file():
+            raise FileNotFoundError(image)
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--output", type=Path, default=Path(".tmp/examples-site"))
+    args = parser.parse_args()
+    args.output.mkdir(parents=True, exist_ok=True)
+    assets = args.output/"assets"
+    assets.mkdir(exist_ok=True)
+    for image in (ROOT/"docs/images").glob("*.png"):
+        shutil.copyfile(image, assets/image.name)
+    readme = (ROOT/"README.md").read_text(encoding="utf-8")
+    progression = readme.split("## The progression", 1)[1].split("## Baseline and calibration reference", 1)[0]
+    page("# Anomalyzer: the progression\n"+progression, args.output/"index.html", "Anomalyzer — numbered examples")
+    page((ROOT/"docs/sparse-data.md").read_text(), args.output/"sparse-data.html", "Anomalyzer — sparse data")
+    print(f"Exported numbered progression and sparse-data guide to {args.output}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/publish_release.py b/scripts/publish_release.py
index 560e006..6296fd2 100644
--- a/scripts/publish_release.py
+++ b/scripts/publish_release.py
@@ -33,7 +33,10 @@ def publish(dist, tag, revision, *, finalize=False):
             raise RuntimeError(found.stderr)
         with tempfile.TemporaryDirectory() as temp:
             notes = Path(temp) / "notes.md"
+            feature_path = Path(__file__).resolve().parents[1] / "docs/releases" / f"{tag[1:]}.md"
+            feature_notes = feature_path.read_text(encoding="utf-8") + "\n\n" if feature_path.is_file() else ""
             notes.write_text(
+                feature_notes +
                 f"Anomalyzer {tag[1:]} — local Python analysis library.\n\n"
                 f"Source revision: `{revision}`.\n\n"
                 "Wheel and source distribution tested on Python 3.11–3.13; "
diff --git a/scripts/verify_distribution.py b/scripts/verify_distribution.py
index e63bb87..df57324 100644
--- a/scripts/verify_distribution.py
+++ b/scripts/verify_distribution.py
@@ -19,6 +19,8 @@ def main():
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument("--dist", type=Path, default=Path("dist"))
     parser.add_argument("--python", default="3.12")
+    parser.add_argument("--from-release", action="store_true",
+                        help="Install the wheel through the exact published URL and hashes")
     args = parser.parse_args()
     dist = args.dist.resolve()
     artifacts = list(dist.glob("*.whl")) + list(dist.glob("*.tar.gz"))
@@ -37,9 +39,13 @@ def main():
             run(["uv", "venv", "--python", args.python, venv], outside, env=env)
             python = venv / ("Scripts/python.exe" if os.name == "nt" else "bin/python")
             scripts = python.parent
-            run(["uv", "pip", "install", "--python", python, "--require-hashes", "-r",
-                 dist / "runtime-dependencies.txt"], outside, env=env)
-            run(["uv", "pip", "install", "--python", python, "--no-deps", artifact], outside, env=env)
+            if args.from_release and artifact.suffix == ".whl":
+                run(["uv", "pip", "install", "--python", python, "--require-hashes", "-r",
+                     dist / "requirements-guard-py312-linux-x86_64.txt"], outside, env=env)
+            else:
+                run(["uv", "pip", "install", "--python", python, "--require-hashes", "-r",
+                     dist / "runtime-dependencies.txt"], outside, env=env)
+                run(["uv", "pip", "install", "--python", python, "--no-deps", artifact], outside, env=env)
             run(["uv", "pip", "check", "--python", python], outside, env=env)
             run([python, "-I", smoke, "--expect-prefix", venv], outside, env=env)
             cli = scripts / ("anomalyzer.exe" if os.name == "nt" else "anomalyzer")
diff --git a/scripts/verify_lambda.sh b/scripts/verify_lambda.sh
index 441fa14..1459d2a 100644
--- a/scripts/verify_lambda.sh
+++ b/scripts/verify_lambda.sh
@@ -5,6 +5,7 @@ image=public.ecr.aws/lambda/python:3.12
 root=$(pwd)
 bundle="$root/.tmp/lambda-consumer"
 mkdir -p "$bundle"
+mkdir -p "$bundle/reports"
 if [ -e "$bundle/package" ]; then
   echo "Use a fresh .tmp/lambda-consumer directory for each verification" >&2
   exit 1
@@ -27,6 +28,8 @@ docker run --rm --platform linux/amd64 --entrypoint /bin/bash \
 docker run --rm --platform linux/amd64 --network none --entrypoint /bin/bash \
   -v "$bundle/package:/var/task:ro" \
   -v "$root/examples/library/guard_smoke.py:/tmp/guard_smoke.py:ro" \
+  -v "$root/examples/library/benchmark_sparse.py:/tmp/benchmark_sparse.py:ro" \
+  -v "$bundle/reports:/reports" \
   -w /var/task "$image" -euc \
-  'PYTHONPATH=/var/task python /tmp/guard_smoke.py --expect-prefix /var/task' \
+  'PYTHONPATH=/var/task python /tmp/guard_smoke.py --expect-prefix /var/task && PYTHONPATH=/var/task python /tmp/benchmark_sparse.py --output /reports/benchmark.json >/dev/null' \
   | tee "$bundle/smoke.json"
diff --git a/src/anomalyzer/__init__.py b/src/anomalyzer/__init__.py
index 5736d02..7bc2f7e 100644
--- a/src/anomalyzer/__init__.py
+++ b/src/anomalyzer/__init__.py
@@ -1,10 +1,11 @@
 """Public, local-only analysis API."""
 from ._version import __version__
-from .contracts import Request, RequestV11, Result, ResultV11
+from .contracts import DataPolicy, DatasetV11, SampleRevision, Request, RequestV11, Result, ResultV11
 from .agent import analyze_relationships, analyze_series, get_case, replay_policy
 from .recipes import analyze, analyze_dataset
 
 __all__ = [
+    "DataPolicy", "DatasetV11", "SampleRevision",
     "Request", "RequestV11", "Result", "ResultV11", "analyze",
     "analyze_dataset", "analyze_series", "analyze_relationships", "get_case",
     "replay_policy",
diff --git a/src/anomalyzer/_version.py b/src/anomalyzer/_version.py
index e2468a0..242d872 100644
--- a/src/anomalyzer/_version.py
+++ b/src/anomalyzer/_version.py
@@ -1,3 +1,3 @@
 """Authoritative distribution and runtime version (read by Hatchling)."""
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"
diff --git a/src/anomalyzer/agent.py b/src/anomalyzer/agent.py
index b237543..d03bc7c 100644
--- a/src/anomalyzer/agent.py
+++ b/src/anomalyzer/agent.py
@@ -3,7 +3,7 @@
 
 from .cases import compose_cases, target_entities
 from .contracts import (
-    ActionPolicy, Case, Context, Dataset, RequestV11, ResultV11, Settings,
+    ActionPolicy, Case, Context, Dataset, DatasetV11, RequestV11, ResultV11, Settings,
 )
 from .recipes import analyze
 
@@ -11,11 +11,18 @@
 def analyze_series(dataset: Dataset | dict | list[float | None],
                    settings: Settings | dict | None = None,
                    context: Context | dict | None = None):
-    """Analyze one series using the frozen schema-1.0 semantics."""
+    """Use legacy semantics for a Dataset; explicit V11 fields select source policies."""
     if isinstance(dataset, list):
         if context is not None:
             raise ValueError("context requires a Dataset rather than a bare values list")
         return analyze(dataset, settings)
+    if isinstance(dataset, DatasetV11) or isinstance(dataset, dict) and any(
+            name in dataset for name in ("data_policy", "coverage", "available_at", "revisions", "rules", "completed_period_rules")):
+        config = Settings.model_validate(settings or {})
+        validated = DatasetV11.model_validate(dataset)
+        return analyze({"schema_version": "1.1", "datasets": [validated.model_dump(mode="json")],
+                        "config": config.model_dump(mode="json", exclude_unset=True),
+                        "context": Context.model_validate(context or {}).model_dump(mode="json")})
     validated = Dataset.model_validate(dataset)
     config = Settings.model_validate(settings or {})
     context = Context.model_validate(context or {})
@@ -53,8 +60,12 @@ def replay_policy(result: ResultV11 | dict,
                     if policy.notification_requires_action else
                     assessment.criterion_met)
     request = RequestV11.model_validate(replayed.resolved_config)
-    replayed.cases = compose_cases(
-        replayed.dataset_results, replayed.relationship_results,
-        target_entities(request))
+    if any(item.resolution for item in replayed.dataset_results):
+        from .sparse_views import view_cases
+        replayed.cases = view_cases(replayed.dataset_results, replayed.relationship_results, request)
+    else:
+        replayed.cases = compose_cases(
+            replayed.dataset_results, replayed.relationship_results,
+            target_entities(request))
     replayed.resolved_config["policy"] = policy.model_dump(mode="json")
     return replayed
diff --git a/src/anomalyzer/alignment.py b/src/anomalyzer/alignment.py
index ad82359..4c791fd 100644
--- a/src/anomalyzer/alignment.py
+++ b/src/anomalyzer/alignment.py
@@ -15,24 +15,28 @@ class PreparedDataset:
 
 
 def prepare_source(dataset: DatasetV11, settings: Settings,
-                   context: Context) -> PreparedDataset:
-    times, values, quality = prepare(dataset, settings, context)
+                   context: Context, budget=None) -> PreparedDataset:
+    from .sparse import normalize, uses_policy
+    explicit = uses_policy(dataset)
+    times, values, quality = (normalize(dataset, settings, context, budget)
+                              if explicit else prepare(dataset, settings, context))
     if dataset.timestamps is None:
         indexes = list(range(len(values)))
     else:
-        if quality["duplicate_count"]:
+        if quality["duplicate_count"] and dataset.data_policy is None:
             raise ValueError(
                 "relationship lineage requires unique source timestamps; duplicates are not supported")
         by_time = {
             timestamp(value, dataset.timezone).isoformat(): index
             for index, value in enumerate(dataset.timestamps)
         }
-        indexes = [by_time[value] for value in times]
+        indexes = ([sample["source_index"] if len(sample["source_indexes"]) <= 1 else -1
+                    for sample in quality["samples"]] if explicit else [by_time.get(value, -1) for value in times])
         # Retain trailing explicitly incomplete periods for relationship
         # alignment, but represent their values as unavailable. The
         # single-series engine still receives the original Dataset and excludes
         # these periods during its own preparation.
-        for item in quality.get("incomplete_periods", []):
+        for item in ([] if explicit else quality.get("incomplete_periods", [])):
             times.append(item["timestamp"])
             values.append(None)
             indexes.append(item["index"])
@@ -46,6 +50,9 @@ def exact_sources(ids: list[str], prepared: dict[str, PreparedDataset]):
     frequencies = {source.dataset.frequency for source in sources}
     if len(frequencies) != 1:
         raise ValueError("related datasets must declare the same cadence")
+    bases = {source.quality.get("coordinate") == "event_index" for source in sources}
+    if len(bases) != 1:
+        raise ValueError("related datasets must share the observation basis")
     entities = {tuple(sorted(source.dataset.entity.items())) for source in sources}
     if len(entities) != 1:
         raise ValueError("related datasets must declare matching entity metadata")
diff --git a/src/anomalyzer/cases.py b/src/anomalyzer/cases.py
index 1952bf0..a558454 100644
--- a/src/anomalyzer/cases.py
+++ b/src/anomalyzer/cases.py
@@ -117,7 +117,10 @@ def compose_cases(dataset_results: list[DatasetResultV11],
             assessment.id for _, _, assessment in items)
         cases.append(Case(
             id=case_id, revision_id=revision_id, entity=entity, event_time=event,
-            detection_time=event, source_availability_cutoff=event,
+            detection_time=max([event, *[assessment.baseline.get("known_input_availability_cutoff", event)
+                                        for _, _, assessment in items]], key=str),
+            source_availability_cutoff=max([event, *[assessment.baseline.get("known_input_availability_cutoff", event)
+                                                    for _, _, assessment in items]], key=str),
             contributing_dataset_ids=datasets,
             contributing_relationship_ids=relationships,
             supporting_evidence=refs, conflicting_evidence=[],
diff --git a/src/anomalyzer/contracts.py b/src/anomalyzer/contracts.py
index bad29c5..4f29b02 100644
--- a/src/anomalyzer/contracts.py
+++ b/src/anomalyzer/contracts.py
@@ -78,7 +78,7 @@ def aligned(self):
         if self.timestamps is None:
             if self.frequency is not None or self.timezone is not None:
                 raise ValueError("frequency and timezone require timestamps")
-        elif self.frequency is None:
+        elif self.frequency is None and getattr(getattr(self, "data_policy", None), "observation_basis", None) != "event_index":
             raise ValueError("timestamps require frequency")
         elif len(self.timestamps) != len(self.values):
             raise ValueError("timestamps and values must have equal length")
@@ -237,11 +237,70 @@ class MaximumRelativeChangeRule(Contract):
 ]
 
 
+class DataPolicy(Contract):
+    """Explicit source semantics; unknown collection is never implicit zero."""
+    missing: Literal["preserve", "zero_for_empty"] = "preserve"
+    measurement: Literal["measurement", "additive_volume"] = "measurement"
+    absence_means_zero: bool = False
+    observation_basis: Literal["calendar", "event_index"] = "calendar"
+    reference_gaps: Literal["same_phase", "restart"] = "same_phase"
+    grid_start: str | None = None
+    grid_end: str | None = None
+    change_max_gap_seconds: float | None = Field(default=None, gt=0)
+    minimum_phase_samples: int = Field(default=2, ge=1, le=10000, strict=True)
+
+    @model_validator(mode="after")
+    def meaningful(self):
+        if (self.missing == "zero_for_empty" or self.absence_means_zero) and self.measurement != "additive_volume":
+            raise ValueError("zero-for-empty requires measurement='additive_volume'")
+        if self.absence_means_zero and self.missing != "zero_for_empty":
+            raise ValueError("absence_means_zero requires missing='zero_for_empty'")
+        if self.observation_basis == "event_index" and (
+                self.missing != "preserve" or self.grid_start or self.grid_end):
+            raise ValueError("event_index preserves real events; zero filling and calendar grids are unavailable")
+        return self
+
+
+Coverage = Literal["complete", "empty", "unknown", "incomplete", "excluded", "revised"]
+
+
+class SampleRevision(Contract):
+    timestamp: str
+    value: float | None
+    available_at: str
+    revision_id: str = Field(min_length=1)
+    coverage: Coverage = "revised"
+
+    @field_validator("value", mode="before")
+    @classmethod
+    def numeric_revision(cls, value):
+        if value is not None and (isinstance(value, bool) or not isinstance(value, (float, int))):
+            raise ValueError("revision values must be numeric or null")
+        return value
+
+
 class DatasetV11(Dataset):
     entity: dict[str, str] = Field(min_length=1)
     rules: list[ExplicitRule] = Field(default_factory=list)
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
+    data_policy: DataPolicy | None = None
+    coverage: list[Coverage] | None = None
+    available_at: list[str | None] | None = None
+    revisions: list[SampleRevision] = Field(default_factory=list)
+
+    @model_validator(mode="after")
+    def source_annotations(self):
+        for name in ("coverage", "available_at"):
+            annotation = getattr(self, name)
+            if annotation is not None and len(annotation) != len(self.values):
+                raise ValueError(f"{name} and values must have equal length")
+        if (self.data_policy or self.coverage or self.available_at or self.revisions) and self.timestamps is None:
+            raise ValueError("source policies/coverage/revisions require original timestamps")
+        if self.data_policy and self.data_policy.observation_basis == "event_index" and self.frequency is not None:
+            raise ValueError("event_index must omit frequency; event position is not elapsed cadence")
+        return self
 
-    @field_validator("rules")
+    @field_validator("rules", "completed_period_rules")
     @classmethod
     def unique_rule_ids(cls, rules):
         ids = [rule.id for rule in rules]
@@ -251,6 +310,7 @@ def unique_rule_ids(cls, rules):
 
 
 class RatioRelationship(Contract):
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
     id: str = Field(min_length=1)
     kind: Literal["ratio"]
     numerator: str = Field(min_length=1)
@@ -262,6 +322,7 @@ class RatioRelationship(Contract):
 
 
 class DifferenceRelationship(Contract):
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
     id: str = Field(min_length=1)
     kind: Literal["difference"]
     minuend: str = Field(min_length=1)
@@ -272,6 +333,7 @@ class DifferenceRelationship(Contract):
 
 
 class NormalizedResidualRelationship(Contract):
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
     id: str = Field(min_length=1)
     kind: Literal["normalized_residual"]
     observed: str = Field(min_length=1)
@@ -283,6 +345,7 @@ class NormalizedResidualRelationship(Contract):
 
 
 class LaggedResponseRelationship(Contract):
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
     id: str = Field(min_length=1)
     kind: Literal["lagged_response"]
     response: str = Field(min_length=1)
@@ -309,6 +372,7 @@ class JointConditionTerm(Contract):
 
 
 class JointConditionRelationship(Contract):
+    completed_period_rules: list[ExplicitRule] = Field(default_factory=list)
     id: str = Field(min_length=1)
     kind: Literal["joint_condition"]
     conditions: list[JointConditionTerm] = Field(min_length=2, max_length=4)
@@ -370,7 +434,7 @@ def unique_ids_and_references(self):
             if len(set(refs)) != len(refs):
                 raise ValueError(
                     f"relationship {relationship.id!r} must reference distinct datasets")
-        total_points = sum(len(dataset.values) for dataset in self.datasets)
+        total_points = sum(len(dataset.values) + len(dataset.revisions) for dataset in self.datasets)
         if total_points > self.config.max_points:
             raise ValueError("request exceeds max_points across all datasets")
         return self
@@ -417,10 +481,16 @@ class LineagePoint(Contract):
     lag: int = 0
     unavailable_reason: str | None = None
     latest_source_cutoff: str
+    source_coverage: dict[str, str] = Field(default_factory=dict)
+    source_values: dict[str, float | None] = Field(default_factory=dict)
+    source_revision_ids: dict[str, str | None] = Field(default_factory=dict)
+    source_available_at: dict[str, str | None] = Field(default_factory=dict)
+    source_input_indexes: dict[str, list[int]] = Field(default_factory=dict)
 
 
 class DatasetResultV11(Contract):
     dataset_id: str
+    resolution: str | None = None
     status: Literal[
         "completed", "partial", "failed", "inapplicable",
         "insufficient_history",
@@ -436,6 +506,7 @@ class DatasetResultV11(Contract):
 
 class RelationshipResult(Contract):
     relationship_id: str
+    resolution: str | None = None
     kind: Literal[
         "ratio", "difference", "normalized_residual", "lagged_response",
         "joint_condition",
@@ -458,6 +529,7 @@ class RelationshipResult(Contract):
 
 class Case(Contract):
     id: str
+    resolution: str | None = None
     revision_id: str
     entity: dict[str, str]
     event_time: str | int
diff --git a/src/anomalyzer/early.py b/src/anomalyzer/early.py
index 7f837aa..ce2c2bd 100644
--- a/src/anomalyzer/early.py
+++ b/src/anomalyzer/early.py
@@ -2,6 +2,9 @@
 from __future__ import annotations
 
 import statistics
+import math
+from bisect import insort
+from datetime import datetime
 from typing import Iterable
 
 from .contracts import (
@@ -39,7 +42,9 @@ def descriptive_assessments(target_id: str, times: list[str | None],
                             values: list[float | None],
                             rules: Iterable[ExplicitRule] = (),
                             policy: ActionPolicy | None = None,
-                            relationship: bool = False) -> list[Assessment]:
+                            relationship: bool = False,
+                            reset_positions=(), change_max_gap_seconds=None,
+                            excluded_positions=()) -> list[Assessment]:
     """Return prefix-invariant facts and simple-reference comparisons.
 
     Missing values break the comparison population. They are intentionally not
@@ -49,14 +54,23 @@ def descriptive_assessments(target_id: str, times: list[str | None],
     result: list[Assessment] = []
     prior: list[float] = []
     previous: float | None = None
+    boundaries = set(reset_positions) | set(excluded_positions) | {i + 1 for i in excluded_positions}
     for index, (observed_at, value) in enumerate(zip(times, values)):
+        if (index in boundaries or change_max_gap_seconds is not None and index and
+                times[index] and times[index-1] and
+                (datetime.fromisoformat(times[index]) - datetime.fromisoformat(times[index-1])).total_seconds() > change_max_gap_seconds):
+            prior, previous = [], None
         if value is None:
             prior = []
             previous = None
             continue
         change = value - previous if previous is not None else None
+        if change is not None and not math.isfinite(change):
+            change = None
         relative = (change / abs(previous)
                     if change is not None and previous != 0 else None)
+        if relative is not None and not math.isfinite(relative):
+            relative = None
         violations = _rule_violations(value, previous, rules)
         if violations:
             classification = "criterion_violation"
@@ -94,8 +108,9 @@ def descriptive_assessments(target_id: str, times: list[str | None],
                 "Two observations do not establish an unusual departure, stable "
                 "baseline, or real-world cause.")
         else:
-            center = statistics.median(prior)
-            low, high = min(prior), max(prior)
+            middle = len(prior)//2
+            center = statistics.mean(prior[middle-1:middle+1]) if len(prior)%2 == 0 else prior[middle]
+            low, high = prior[0], prior[-1]
             outside = value < low or value > high
             width = high - low
             distance = low - value if value < low else value - high if value > high else 0.0
@@ -120,6 +135,13 @@ def descriptive_assessments(target_id: str, times: list[str | None],
                 "does not establish a real-world anomaly or cause.")
         action = bool(
             criterion_met and policy.allow_explicit_rules)
+        if strength is not None and not math.isfinite(strength):
+            strength = None
+        unavailable_rules = [dict(id=rule.id, reason="no_comparable_prior" if previous is None else "relative_change_from_zero")
+            for rule in rules if rule.kind != "acceptable_range" and
+            (previous is None or rule.kind == "maximum_relative_change" and previous == 0)]
+        if unavailable_rules:
+            baseline["unavailable_rules"] = unavailable_rules
         result.append(Assessment(
             id=f"{target_id}:assessment:{index}:early",
             target_id=target_id, index=index, timestamp=observed_at,
@@ -138,7 +160,7 @@ def descriptive_assessments(target_id: str, times: list[str | None],
                 "Values share the declared target units and entity.",
                 "No interpolation, resampling, or implicit unit conversion was performed.",
             ], established=established, not_established=not_established))
-        prior.append(value)
+        insort(prior, value)
         previous = value
     return result
 
diff --git a/src/anomalyzer/mcp_views.py b/src/anomalyzer/mcp_views.py
index 4f85e63..627a682 100644
--- a/src/anomalyzer/mcp_views.py
+++ b/src/anomalyzer/mcp_views.py
@@ -67,6 +67,12 @@ def _quality(item):
         "regular": source.get("regular"),
         "latest_evidence_state": source.get("latest_evidence_state"),
         "error": item.get("error"),
+        **({"data_policy": source["data_policy"], "coverage_counts": source["coverage_counts"],
+            "grid_count": source["grid_count"], "usable_model_count": source["usable_model_count"],
+            "reported_measurement_count": source["reported_measurement_count"],
+            "zero_filled_count": source["zero_filled_count"], "observation_basis": source["coordinate"]}
+           if "data_policy" in source else {}),
+        **({"resolution": item["resolution"]} if item.get("resolution") else {}),
     }
 
 
@@ -117,7 +123,7 @@ def _interval_text(interval):
 
 def _quality_headline(target, item, quality):
     if quality["latest_evidence_state"] == "insufficient_evidence":
-        return f"{target}: latest relationship evidence is unavailable."
+        return f"{target}: latest numerical evidence is unavailable."
     if quality["incomplete_count"]:
         return (f"{target}: {quality['incomplete_count']} incomplete source "
                 "period(s) were excluded from finalized scoring.")
@@ -139,17 +145,19 @@ def build_findings(result):
             target = matches[0][1] if matches else None
             findings.append({
                 "id": case["id"], "kind": "case", "target_id": target,
+                **({"resolution": case["resolution"]} if case.get("resolution") else {}),
                 "entity": case["entity"], "event_time": case["event_time"],
                 "headline": _case_headline(case, matches, definitions),
                 "maturity": case["maturity"],
                 "severity": case["severity"],
                 "readiness": (_readiness(next(item for _, name, item in _targets(result)
-                                              if name == target))[0]
+                                              if name == target and item.get("resolution") == case.get("resolution")))[0]
                               if target else "not_assessed"),
                 "note": "A case is numerical evidence, not an incident or permission to act.",
                 "_order": order, "_priority": 0,
             })
     for kind, target, item in _targets(result):
+        previous_findings = len(findings)
         readiness, _ = _readiness(item)
         for episode in item["observations"]:
             findings.append({
@@ -202,7 +210,7 @@ def build_findings(result):
         if _quality_issue(item):
             quality = _quality(item)
             findings.append({
-                "id": f"quality:{kind}:{target}", "kind": "quality",
+                "id": f"quality:{kind}:{target}" + (f":{item['resolution']}" if item.get("resolution") else ""), "kind": "quality",
                 "target_id": target, "entity": definitions[target]["entity"],
                 "event_time": item["data_quality"].get("window", [None, None])[-1]
                               if item["data_quality"].get("window") else None,
@@ -215,7 +223,11 @@ def build_findings(result):
                     or quality["missing_count"] or quality["incomplete_count"]
                     else 3),
             })
-    findings.sort(key=lambda item: (item["_priority"], -item["_order"], item["id"]))
+        if item.get("resolution"):
+            for finding in findings[previous_findings:]:
+                finding["resolution"] = item["resolution"]
+    findings.sort(key=lambda item: (item["_priority"], -item["_order"],
+        item["kind"] == "quality" and not item["id"].startswith("quality:relationship:"), item["id"]))
     for item in findings:
         del item["_order"]
         del item["_priority"]
@@ -289,7 +301,7 @@ def get_evidence(result, finding_id, cursor=0, limit=50):
         references = set(record["evidence_refs"])
         targets = {target for _, target, _ in matches}
         for kind, target, item in _targets(result):
-            if target in targets:
+            if target in targets and item.get("resolution") == record.get("resolution"):
                 related.append((kind, target, item))
                 if kind == "relationship":
                     indexes = {assessment["index"] for source_kind, name, assessment
@@ -299,7 +311,7 @@ def get_evidence(result, finding_id, cursor=0, limit=50):
     else:
         kind, target, item = next((kind, target, item)
                                   for kind, target, item in _targets(result)
-                                  if target == finding["target_id"])
+                                  if target == finding["target_id"] and item.get("resolution") == finding.get("resolution"))
         related = [(kind, target, item)]
         if finding["kind"] == "pattern":
             record = next(pattern for pattern in item["anomaly_patterns"]
@@ -321,6 +333,9 @@ def get_evidence(result, finding_id, cursor=0, limit=50):
             record = {
                 "status": item["status"], "data_quality": _quality(item),
                 "transformations": source_quality.get("transformations", []),
+                "source_samples_preview": source_quality.get("samples", [])[-5:],
+                "aggregation_preview": (source_quality.get("aggregation", [])[:5]
+                    if isinstance(source_quality.get("aggregation", []), list) else source_quality.get("aggregation")),
                 "incomplete_period_count": len(incomplete),
                 "incomplete_periods_preview": incomplete[:5],
                 "error": item.get("error"),
diff --git a/src/anomalyzer/orchestrator.py b/src/anomalyzer/orchestrator.py
index 32ecfcb..4a7b268 100644
--- a/src/anomalyzer/orchestrator.py
+++ b/src/anomalyzer/orchestrator.py
@@ -18,7 +18,24 @@
 )
 from .recipes import _namespace_result, analyze_dataset
 from .registry import versions
-from .relationships import derive
+from .relationships import derive, _source_ids
+from .validation import resolve_reset_points
+from .sparse import analyze_sparse, active_resets
+
+
+def annotate_availability(assessments, times, samples):
+    """Conservative known-input cutoff; not a measured first-detection latency."""
+    if not any(row.get("available_at") and (row.get("value") is not None or row.get("coverage") == "excluded") for row in samples):
+        return
+    cutoffs, latest = [], None
+    for observed_at, row in zip(times, samples):
+        if row.get("value") is not None or row.get("coverage") == "excluded":
+            latest = max(filter(None, (latest, observed_at, row.get("available_at"))))
+        cutoffs.append(max(filter(None, (latest, observed_at))))
+    for assessment in assessments:
+        if assessment.index < len(cutoffs):
+            assessment.baseline["known_input_availability_cutoff"] = cutoffs[assessment.index]
+            assessment.assumptions.append("Availability cutoff is conservative across known prior inputs; unknown arrival history cannot establish delivery-time detection performance.")
 
 
 def _canonical_request(request: RequestV11) -> dict:
@@ -47,6 +64,8 @@ def _derived_analysis_view(derived):
     analyze the later derived series without treating those points as zero.
     Internal or trailing gaps remain visible and make the engine inapplicable.
     """
+    if hasattr(derived, "data_policy"):
+        return derived, 0
     offset = 0
     while offset < len(derived.values) and derived.values[offset] is None:
         offset += 1
@@ -82,40 +101,53 @@ def _shift_result_indexes(result, offset):
     return result
 
 
-def analyze_relationships(request: RequestV11 | dict) -> ResultV11:
+def analyze_relationships(request: RequestV11 | dict, *, _budget=None) -> ResultV11:
     request = RequestV11.model_validate(request)
+    if request.config.recipe == "multi-resolution-v1":
+        from .sparse_views import analyze_multi_request
+        return analyze_multi_request(request, _budget)
     if request.config.recipe not in (
             "seasonal-residual-v1", "adaptive-seasonal-v1"):
         raise ValueError(
             "schema 1.1 relationships require seasonal-residual-v1 or "
             "adaptive-seasonal-v1")
     started = time.perf_counter()
-    budget = RuntimeBudget(request.config.max_runtime_seconds)
+    budget = _budget or RuntimeBudget(request.config.max_runtime_seconds)
     canonical = _canonical_request(request)
     fingerprint = hashlib.sha256(json.dumps(
         canonical, sort_keys=True, separators=(",", ":"),
         allow_nan=False).encode()).hexdigest()
     prepared: dict[str, PreparedDataset] = {}
     dataset_results = []
+    prepared_points = 0
     for dataset in sorted(request.datasets, key=lambda item: item.id):
         try:
             budget.check()
-            source = prepare_source(dataset, request.config, request.context)
+            source = prepare_source(dataset, request.config, request.context, budget)
+            prepared_points += len(source.values) + len(dataset.revisions)
+            if prepared_points > request.config.max_points:
+                raise ValueError("expanded request exceeds shared max_points across all datasets")
             prepared[dataset.id] = source
-            result = analyze_dataset(
-                dataset, request.config, request.context, budget=budget)
+            result = (analyze_sparse(dataset, request.config, request.context, budget,
+                       (source.times, source.values, source.quality))
+                      if "data_policy" in source.quality and request.config.recipe != "multi-resolution-v1"
+                      else analyze_dataset(dataset, request.config, request.context, budget=budget))
             _namespace_result(result, f"dataset:{dataset.id}")
             assessments = descriptive_assessments(
                 dataset.id,
-                source.times[:source.quality["observation_count"]],
-                source.values[:source.quality["observation_count"]], dataset.rules,
-                request.policy)
+                source.times[:source.quality.get("grid_count", source.quality["observation_count"])],
+                source.quality.get("rule_values", source.values)[:source.quality.get("grid_count", source.quality["observation_count"])], dataset.rules,
+                request.policy,
+                reset_positions=active_resets(request.config.reset_points, source.times[:source.quality.get("grid_count", source.quality["observation_count"])], dataset.timezone, request.context.as_of),
+                change_max_gap_seconds=dataset.data_policy.change_max_gap_seconds if dataset.data_policy else None,
+                excluded_positions=[s["index"] for s in source.quality.get("samples", []) if s["coverage"] == "excluded"])
             assessments.extend(calibrated_assessments(
                 dataset.id, result.methods, request.policy))
-            if request.config.recipe == "adaptive-seasonal-v1":
+            if request.config.recipe == "adaptive-seasonal-v1" or "data_policy" in source.quality:
                 assessments.extend(location_pattern_assessments(
                     dataset.id, result.methods, result.anomaly_patterns,
                     request.policy))
+            annotate_availability(assessments, source.times, source.quality.get("samples", []))
             dataset_results.append(DatasetResultV11(
                 dataset_id=dataset.id, status=result.status,
                 data_quality=result.data_quality, methods=result.methods,
@@ -130,6 +162,11 @@ def analyze_relationships(request: RequestV11 | dict) -> ResultV11:
         definition = relationship.model_dump(mode="json")
         try:
             budget.check()
+            definitions = {d.id: d for d in request.datasets}
+            if any(definitions[name].timestamps is None for name in _source_ids(relationship)):
+                raise ValueError("exact relationship alignment requires timestamped datasets")
+            if len({definitions[name].frequency for name in _source_ids(relationship)}) != 1:
+                raise ValueError("related datasets must declare the same cadence")
             derived, lineage, quality = derive(relationship, prepared)
             analysis_view, analysis_offset = _derived_analysis_view(derived)
             result = analyze_dataset(
@@ -139,14 +176,21 @@ def analyze_relationships(request: RequestV11 | dict) -> ResultV11:
             quality["analysis_offset"] = analysis_offset
             assessments = descriptive_assessments(
                 relationship.id, derived.timestamps or [], derived.values,
-                relationship.rules, request.policy, relationship=True)
+                relationship.rules, request.policy, relationship=True,
+                reset_positions=active_resets(request.config.reset_points, derived.timestamps or [], derived.timezone, request.context.as_of),
+                change_max_gap_seconds=getattr(getattr(derived, "data_policy", None), "change_max_gap_seconds", None),
+                excluded_positions=[i for i, state in enumerate(getattr(derived, "coverage", None) or []) if state == "excluded"])
             assessments.extend(calibrated_assessments(
                 relationship.id, result.methods, request.policy,
                 relationship=True))
-            if request.config.recipe == "adaptive-seasonal-v1":
+            if request.config.recipe == "adaptive-seasonal-v1" or hasattr(derived, "data_policy"):
                 assessments.extend(location_pattern_assessments(
                     relationship.id, result.methods, result.anomaly_patterns,
                     request.policy, relationship=True))
+            annotate_availability(assessments, [p.timestamp for p in lineage],
+                [dict(value=p.value, available_at=p.latest_source_cutoff if any(p.source_available_at.values()) else None) for p in lineage])
+            quality["model_eligibility"] = {key: result.data_quality.get(key)
+                for key in ("coordinate", "usable_model_count", "coverage_counts")}
             if quality["available_count"] == 0 or quality["latest_evidence_state"] == "insufficient_evidence":
                 status = "insufficient_evidence"
             elif quality["unavailable_count"] and result.status == "inapplicable":
diff --git a/src/anomalyzer/recipes.py b/src/anomalyzer/recipes.py
index 31d2f27..7d749a5 100644
--- a/src/anomalyzer/recipes.py
+++ b/src/anomalyzer/recipes.py
@@ -45,6 +45,12 @@ def _analyze_seasonal(request: Request) -> Result:
 def analyze_dataset(dataset: Dataset, settings: Settings, context,
                     budget: RuntimeBudget | None = None) -> Result:
     """Emit causal seasonal evidence, and flag only fully calibrated scores."""
+    from .sparse import analyze_sparse, uses_policy
+    if uses_policy(dataset):
+        if settings.recipe == "multi-resolution-v1":
+            from .sparse_views import analyze_sparse_views
+            return analyze_sparse_views(dataset, settings, context, budget)
+        return analyze_sparse(dataset, settings, context, budget)
     if settings.recipe == "multi-resolution-v1":
         return analyze_multi_resolution(dataset, settings, context, budget)
     start = time.perf_counter()
diff --git a/src/anomalyzer/relationships.py b/src/anomalyzer/relationships.py
index 2fb56b3..3403126 100644
--- a/src/anomalyzer/relationships.py
+++ b/src/anomalyzer/relationships.py
@@ -2,9 +2,10 @@
 from __future__ import annotations
 
 import math
+from dataclasses import replace
 
 from .alignment import PreparedDataset, exact_sources
-from .contracts import Dataset, LineagePoint, Relationship
+from .contracts import DataPolicy, Dataset, DatasetV11, LineagePoint, Relationship
 
 
 def _compare(value: float, operator: str, threshold: float) -> bool:
@@ -31,6 +32,8 @@ def derive(relationship: Relationship,
            prepared: dict[str, PreparedDataset]) -> tuple[Dataset, list[LineagePoint], dict]:
     ids = _source_ids(relationship)
     sources = exact_sources(ids, prepared)
+    # Exclusion from a model does not erase a genuine measured value.
+    sources = [replace(source, values=source.quality.get("rule_values", source.values)) for source in sources]
     by_id = {source.dataset.id: source for source in sources}
     times = sources[0].times
     if any(value is None for value in times):
@@ -76,6 +79,7 @@ def derive(relationship: Relationship,
     values: list[float | None] = []
     lineage: list[LineagePoint] = []
     unavailable_count = 0
+    excluded_indexes = set()
     for index, observed_at in enumerate(times):
         used: list[tuple[PreparedDataset, int]] = []
         unavailable = None
@@ -146,8 +150,10 @@ def derive(relationship: Relationship,
                            for value, condition in zip(source_values, relationship.conditions)]
                 value = float(all(matches) if relationship.combine == "all" else any(matches))
         if value is not None and not math.isfinite(value):
-            raise ValueError(f"derived nonfinite value at aligned index {index}")
+            value, unavailable = None, "nonfinite_derived_value"
         values.append(value)
+        if any("samples" in source.quality and source.quality["samples"][i]["coverage"] == "excluded" for source, i in used):
+            excluded_indexes.add(index)
         unavailable_count += unavailable is not None
         source_indexes = {source.dataset.id: source.source_indexes[source_index]
                           for source, source_index in used}
@@ -161,11 +167,34 @@ def derive(relationship: Relationship,
             formula=formula, parameters=parameters, units=relationship.units,
             lag=lag, unavailable_reason=unavailable,
             latest_source_cutoff=cutoff))
+        lineage[-1].source_values = {source.dataset.id: source.values[i] for source, i in used}
+        lineage[-1].source_coverage = {source.dataset.id:
+            source.quality["samples"][i]["coverage"] if "samples" in source.quality else
+            "complete" if source.values[i] is not None else "unknown_or_incomplete" for source, i in used}
+        lineage[-1].source_revision_ids = {source.dataset.id:
+            source.quality["samples"][i]["revision_id"] if "samples" in source.quality else None for source, i in used}
+        lineage[-1].source_available_at = {source.dataset.id:
+            source.quality["samples"][i]["available_at"] if "samples" in source.quality else None for source, i in used}
+        lineage[-1].source_input_indexes = {source.dataset.id:
+            source.quality["samples"][i]["source_indexes"] if "samples" in source.quality else [source.source_indexes[i]] for source, i in used}
+        known = [lineage[-1].source_available_at[source.dataset.id] for source, i in used
+                 if source.values[i] is not None and lineage[-1].source_available_at[source.dataset.id]]
+        lineage[-1].latest_source_cutoff = max([cutoff, *known])
     first = sources[0].dataset
-    derived = Dataset(
-        id=relationship.id, timestamps=times, values=values,
-        frequency=first.frequency, units=relationship.units,
-        entity=first.entity)
+    if any("data_policy" in source.quality for source in sources) or any(v is None for v in values):
+        basis = "event_index" if sources[0].quality.get("coordinate") == "event_index" else "calendar"
+        policies = [getattr(source.dataset, "data_policy", None) or DataPolicy() for source in sources]
+        separation = [policy.change_max_gap_seconds for policy in policies if policy.change_max_gap_seconds]
+        derived = DatasetV11(id=relationship.id, timestamps=times, values=values,
+            frequency=first.frequency, units=relationship.units, entity=first.entity, timezone=first.timezone,
+            data_policy=DataPolicy(observation_basis=basis,
+                reference_gaps="restart" if any(p.reference_gaps == "restart" for p in policies) else "same_phase",
+                minimum_phase_samples=max(p.minimum_phase_samples for p in policies),
+                change_max_gap_seconds=min(separation) if separation else None),
+            coverage=["excluded" if i in excluded_indexes else "complete" if v is not None else "unknown" for i, v in enumerate(values)])
+    else:
+        derived = Dataset(id=relationship.id, timestamps=times, values=values,
+            frequency=first.frequency, units=relationship.units, entity=first.entity, timezone=first.timezone)
     quality = {
         "alignment": "exact", "cadence": first.frequency,
         "aligned_count": len(times), "available_count": len(times) - unavailable_count,
diff --git a/src/anomalyzer/sparse.py b/src/anomalyzer/sparse.py
new file mode 100644
index 0000000..47c4abf
--- /dev/null
+++ b/src/anomalyzer/sparse.py
@@ -0,0 +1,404 @@
+"""Schema-1.1 source policies and causal sparse references; never impute a measurement.
+
+Calendar coordinates remain dense only inside an explicitly bounded input grid.
+Event coordinates retain actual timestamps and represent comparable events, not time.
+"""
+from __future__ import annotations
+
+from collections import Counter
+from datetime import timedelta
+import hashlib
+import json
+import math
+import re
+import statistics
+import time
+import uuid
+
+from .budget import RuntimeBudget
+from .contracts import Context, DataPolicy, Dataset, DatasetV11, MethodResult, Result
+from .evidence import anomaly_patterns, episodes
+from .registry import versions
+from .robust import extreme_indices
+from .validation import resolve_reset_points, timestamp
+
+
+def cadence_seconds(frequency):
+    number, unit = re.fullmatch(r"([1-9][0-9]*)(s|min|h|d|w)", frequency).groups()
+    return int(number) * {"s": 1, "min": 60, "h": 3600, "d": 86400, "w": 604800}[unit]
+
+
+def uses_policy(ds):
+    if not isinstance(ds, DatasetV11) or ds.timestamps is None:
+        return False
+    if ds.data_policy or ds.coverage is not None or ds.available_at is not None or ds.revisions or any(v is None for v in ds.values) or "incomplete" in (ds.period_statuses or []):
+        return True
+    times = sorted(timestamp(t, ds.timezone) for t in ds.timestamps)
+    step = timedelta(seconds=cadence_seconds(ds.frequency))
+    return any(b - a not in (timedelta(0), step) for a, b in zip(times, times[1:]))
+
+
+def normalize(ds, cfg, context, budget=None):
+    """Return the explicit grid, eligible model values and inspectable provenance."""
+    budget = budget or RuntimeBudget(cfg.max_runtime_seconds)
+    policy = ds.data_policy or DataPolicy()
+    context = Context.model_validate(context or {})
+    cutoff = timestamp(context.as_of, ds.timezone) if context.as_of else None
+    rows = {}
+    statuses = ds.period_statuses or ["complete"] * len(ds.values)
+    for index, (text, value, status) in enumerate(zip(ds.timestamps, ds.values, statuses)):
+        budget.check()
+        observed_at = timestamp(text, ds.timezone)
+        coverage = ds.coverage[index] if ds.coverage else (
+            "incomplete" if status == "incomplete" else
+            "complete" if value is not None else "empty" if policy.absence_means_zero else "unknown")
+        if status == "incomplete" and coverage != "incomplete":
+            raise ValueError("incomplete period_status conflicts with coverage")
+        available = timestamp(ds.available_at[index], ds.timezone) if ds.available_at and ds.available_at[index] else None
+        if available and available < observed_at:
+            raise ValueError("available_at cannot precede event time")
+        if cutoff and observed_at > cutoff:
+            continue
+        if observed_at in rows:
+            if cfg.duplicate_policy == "error" or policy.observation_basis == "event_index":
+                raise ValueError("duplicate timestamps require declared duplicate_policy; event_index needs unique event timestamps")
+            previous = rows[observed_at]
+            if coverage != "complete" or previous["coverage"] != "complete" or ds.available_at or ds.revisions:
+                raise ValueError("duplicate aggregation requires complete unrevisioned values")
+            previous["duplicates"].append(value)
+            parts = previous["duplicates"]
+            combined = None if any(v is None for v in parts) else math.fsum(parts)
+            if combined is not None and cfg.duplicate_policy == "mean":
+                combined /= len(parts)
+            if combined is not None and not math.isfinite(combined):
+                raise ValueError("duplicate aggregation produced nonfinite value")
+            previous["original_value"] = combined
+            previous.setdefault("source_indexes", [previous["source_index"]]).append(index)
+            continue
+        rows[observed_at] = dict(source_index=index, original_value=value, coverage=coverage,
+                                 period_status=status, available_at=available, revision_id=None, duplicates=[value])
+    revision_keys = set()
+    for revision in ds.revisions:
+        budget.check()
+        observed_at = timestamp(revision.timestamp, ds.timezone)
+        available = timestamp(revision.available_at, ds.timezone)
+        key = (observed_at, available)
+        if key in revision_keys:
+            raise ValueError("revisions need unique event-time/available-at pairs")
+        revision_keys.add(key)
+        if available < observed_at:
+            raise ValueError("revision available_at cannot precede event time")
+        if cutoff and (observed_at > cutoff or available > cutoff):
+            continue
+        previous = rows.get(observed_at)
+        if previous is not None and previous["available_at"] and previous["available_at"] > available and (not cutoff or previous["available_at"] <= cutoff):
+            continue
+        rows[observed_at] = dict(source_index=previous["source_index"] if previous else -1,
+                                 original_value=revision.value, coverage=revision.coverage,
+                                 period_status="incomplete" if revision.coverage == "incomplete" else "complete",
+                                 available_at=available, revision_id=revision.revision_id, duplicates=[])
+    times = sorted(rows)
+    if policy.observation_basis == "calendar":
+        known_times = [t for t in times if not cutoff or not rows[t]["available_at"] or rows[t]["available_at"] <= cutoff]
+        step = timedelta(seconds=cadence_seconds(ds.frequency))
+        start = timestamp(policy.grid_start, ds.timezone) if policy.grid_start else (known_times[0] if known_times else None)
+        end = timestamp(policy.grid_end, ds.timezone) if policy.grid_end else (known_times[-1] if known_times else None)
+        if start is None or end is None:
+            grid = []
+        else:
+            if end < start:
+                raise ValueError("grid_end precedes grid_start")
+            if (end - start) % step:
+                raise ValueError("grid_end is off the declared cadence grid")
+            if any(t < start or t > end or (t - start) % step for t in known_times):
+                raise ValueError("timestamp outside or off the declared cadence grid")
+            if cutoff:
+                end = min(end, cutoff)
+            count = max(0, (end - start) // step + 1)
+            if count + len(ds.revisions) > cfg.max_points:
+                raise ValueError("expanded grid and revisions exceed max_points")
+            grid = [start + i * step for i in range(count)]
+    else:
+        if len(times) + len(ds.revisions) > cfg.max_points:
+            raise ValueError("events and revisions exceed max_points")
+        grid = [t for t in times if not cutoff or not rows[t]["available_at"] or rows[t]["available_at"] <= cutoff]
+    samples, values, rule_values = [], [], []
+    for index, observed_at in enumerate(grid):
+        budget.check()
+        row = rows.get(observed_at, dict(source_index=-1, original_value=None,
+                       coverage="empty" if policy.absence_means_zero else "unknown",
+                       available_at=None, revision_id=None))
+        state, original = row["coverage"], row["original_value"]
+        if row["available_at"] and cutoff and row["available_at"] > cutoff:
+            state, original = "unknown", None
+            reason = "not_available_as_of"
+        else:
+            reason = None
+        if state == "empty" and original not in (None, 0):
+            raise ValueError("empty coverage conflicts with nonzero observed value")
+        if state in ("complete", "revised") and original is None:
+            state = "unknown"
+        value = original if state in ("complete", "revised") else None
+        if state == "empty" and policy.missing == "zero_for_empty":
+            value, reason = 0.0, "confirmed_empty_additive_volume"
+        if reason is None and value is None:
+            reason = {"empty": "empty_preserved_as_missing", "unknown": "collection_unknown",
+                      "incomplete": "period_incomplete", "excluded": "explicit_model_exclusion"}[state]
+        values.append(value)
+        rule_values.append(original if state == "excluded" else value)
+        samples.append(dict(index=index, source_index=row["source_index"], original_value=original,
+                            source_indexes=row.get("source_indexes", [row["source_index"]] if row["source_index"] >= 0 else []),
+                            origin="duplicate_aggregate" if "source_indexes" in row else "revision" if row["revision_id"] else "supplied" if observed_at in rows else "omitted",
+                            original_period_status=row.get("period_status"), original_coverage=row["coverage"],
+                            value=value, coverage=state, reason=reason,
+                            available_at=row["available_at"].isoformat() if row["available_at"] else None,
+                            revision_id=row["revision_id"]))
+    quality = dict(input_count=len(ds.values), grid_count=len(grid),
+        observation_count=sum(v is not None for v in rule_values),
+        reported_measurement_count=sum(s["original_value"] is not None and s["coverage"] in ("complete", "revised", "excluded") for s in samples),
+        complete_count=sum(v is not None for v in values),
+        incomplete_count=sum(s["coverage"] == "incomplete" for s in samples),
+        incomplete_periods=[dict(index=s["index"], timestamp=grid[s["index"]].isoformat(),
+                                observed=s["original_value"], period_status="incomplete")
+                            for s in samples if s["coverage"] == "incomplete"],
+        missing_count=sum(v is None for v in values), duplicate_count=len(ds.values)-len({timestamp(t, ds.timezone) for t in ds.timestamps}),
+        regular=True, transformations=["bounded declared grid; no interpolation or carry-forward"],
+        timestamp_normalization="UTC", coordinate=policy.observation_basis,
+        regularity_basis="declared_cadence" if policy.observation_basis == "calendar" else "explicit_comparable_events",
+        window=[grid[0].isoformat(), grid[-1].isoformat()] if grid else None,
+        units=ds.units, entity=ds.entity, data_policy=policy.model_dump(mode="json"),
+        samples=samples, rule_values=rule_values,
+        coverage_counts=dict(Counter(s["coverage"] for s in samples)),
+        zero_filled_count=sum(s["reason"] == "confirmed_empty_additive_volume" for s in samples),
+        omitted_count=sum(s["origin"] == "omitted" for s in samples),
+        usable_model_count=sum(v is not None for v in values),
+        measured_value_count=sum(v is not None for v in rule_values),
+        duplicate_sources=[dict(timestamp=t.isoformat(), source_indexes=row["source_indexes"], original_values=row["duplicates"])
+                           for t, row in rows.items() if "source_indexes" in row],
+        latest_evidence_state="insufficient_evidence" if values and values[-1] is None else "available",
+        arrival_semantics="supplied availability timestamps; unannotated samples are event-time snapshots" if any(s["available_at"] and (s["value"] is not None or s["coverage"] == "excluded") for s in samples) else
+                          "event-time snapshot only; historical delivery timing is unknown")
+    return [t.isoformat() for t in grid], values, quality
+
+
+def analyze_sparse(ds, cfg, context, budget=None, prepared=None):
+    """Frozen training phase medians, eligible calibration counts, gap-broken patterns."""
+    started = time.perf_counter()
+    budget = budget or RuntimeBudget(cfg.max_runtime_seconds)
+    context = Context.model_validate(context or {})
+    times, values, quality = prepared or normalize(ds, cfg, context, budget)
+    policy = ds.data_policy or DataPolicy()
+    resets = active_resets(cfg.reset_points, times, ds.timezone, context.as_of)
+    if policy.observation_basis == "calendar" and (cfg.recipe == "adaptive-seasonal-v1" or cfg.trend not in ("none", "auto")):
+        return analyze_contiguous(ds, cfg, context, budget, (times, values, quality), resets)
+    method = MethodResult(id="sparse_seasonal", status="insufficient_history",
+        applicability="eligible observations on the original grid; frozen training phase reference",
+        parameters={**cfg.model_dump(), "resolved_reset_positions": resets,
+                    "profile_revision": "sparse-frozen-phase-v1",
+                    "observation_basis": policy.observation_basis, "reference_updates": "frozen_until_explicit_reset"},
+        diagnostics={"segments": [], "unavailable": [], "pattern_boundaries": resets.copy()})
+    unsupported = ("event_index requires season_length=1, trend=none/auto and a single-resolution seasonal recipe"
+        if policy.observation_basis == "event_index" and
+        (cfg.season_length != 1 or cfg.trend not in ("none", "auto") or cfg.recipe != "seasonal-residual-v1")
+        else "sparse references currently require seasonal-residual-v1 and trend=none/auto; other recipes need complete segments"
+        if cfg.recipe != "seasonal-residual-v1" or cfg.trend not in ("none", "auto") else None)
+    patterns, observations = [], []
+    boundaries = sorted(set([0, *resets, len(values)] + (
+        [i + 1 for i, value in enumerate(values[:-1]) if value is None]
+        if policy.reference_gaps == "restart" else [])))
+    try:
+        if unsupported:
+            method.status, method.applicability = "inapplicable", unsupported
+        else:
+            for segment_number, (start, stop) in enumerate(zip(boundaries, boundaries[1:])):
+                budget.check()
+                training, calibration = [], []
+                phase_values, phase_indexes, phase_expected = {}, {}, {}
+                center = scale = None
+                evidence = []
+                segment = dict(number=segment_number, start_position=start, end_position=stop-1,
+                    training_origin=times[start] if start < len(times) else None,
+                    training_indexes=[], calibration_indexes=[], phase_references=[],
+                    status="insufficient_history", trend="none (explicit sparse frozen reference)")
+                for index in range(start, stop):
+                    budget.check()
+                    value = values[index]
+                    phase = (index - start) % cfg.season_length
+                    if len(training) < cfg.training_size:
+                        if value is not None:
+                            training.append(index)
+                            phase_values.setdefault(phase, []).append(value)
+                            phase_indexes.setdefault(phase, []).append(index)
+                        if len(training) == cfg.training_size:
+                            phase_expected = {p: (safe_median(v) if cfg.outlier_handling == "robust" else statistics.mean(v)) for p, v in phase_values.items()
+                                              if len(v) >= policy.minimum_phase_samples}
+                            segment["training_indexes"] = training.copy()
+                            segment["phase_references"] = [dict(phase=p, expected=phase_expected.get(p),
+                                indexes=phase_indexes[p]) for p, v in sorted(phase_values.items())]
+                        elif value is None:
+                            method.diagnostics["unavailable"].append(dict(index=index, reason="missing_training_sample"))
+                        continue
+                    expected = phase_expected.get(phase)
+                    if value is None or expected is None:
+                        method.diagnostics["unavailable"].append(dict(index=index,
+                            reason="missing_observation" if value is None else "no_training_support_for_calendar_phase",
+                            modeled_expectation=expected))
+                        continue
+                    residual = value - expected
+                    if not math.isfinite(residual):
+                        method.diagnostics["unavailable"].append(dict(index=index, reason="nonfinite_residual"))
+                        continue
+                    maturity, z, triggers = "reference_only", None, []
+                    count = len(calibration)
+                    if count < cfg.calibration_size:
+                        if count >= 3:
+                            raw = [item[1] for item in calibration]
+                            z = (residual - statistics.mean(raw)) / max(statistics.stdev(raw), cfg.scale_floor)
+                            maturity = "provisional"
+                        calibration.append((index, residual))
+                        segment["calibration_indexes"].append(index)
+                    else:
+                        if center is None:
+                            raw = [item[1] for item in calibration]
+                            offsets, _ = extreme_indices(raw, cfg.scale_floor) if cfg.outlier_handling == "robust" else ([], None)
+                            retained = [v for offset, v in enumerate(raw) if offset not in offsets]
+                            if len(retained) < 3:
+                                retained, offsets = raw, []
+                            segment["excluded_calibration_indexes"] = [calibration[i][0] for i in offsets]
+                            raw = retained
+                            center, scale = statistics.mean(raw), max(statistics.stdev(raw), cfg.scale_floor)
+                            segment.update(calibration_center=center, calibration_scale=scale,
+                                           retained_calibration_samples=len(raw),
+                                           calibration_cutoff=times[calibration[-1][0]])
+                        count = segment["retained_calibration_samples"]
+                        z = (residual - center) / scale
+                        maturity = "calibrated"
+                        if abs(z) > cfg.point_threshold:
+                            triggers = ["point"]
+                        segment["status"] = "calibrated"
+                    if z is not None and not math.isfinite(z):
+                        method.diagnostics["unavailable"].append(dict(index=index, reason="nonfinite_standardized_residual"))
+                        continue
+                    relative = residual / abs(expected) if expected else None
+                    evidence.append(dict(id=f"sparse_seasonal:{index}", method=method.id, index=index,
+                        timestamp=times[index], training_cutoff=times[training[-1]], observed=value,
+                        expected=expected, residual=residual,
+                        relative_deviation=relative if relative is None or math.isfinite(relative) else None,
+                        standardized_residual=z, signal_maturity=maturity, calibration_samples=count,
+                        triggers=triggers, score_semantics="frozen sparse phase residual; provisional evidence cannot trigger; not a probability",
+                        excluded_from_model=False, model_value=value, exclusion_reason=None, reference_action="use_observed"))
+                segment["training_indexes"] = training
+                method.diagnostics["segments"].append(segment)
+                method.evidence.extend(evidence)
+                temporary = method.model_copy(update={"evidence": evidence})
+                # Separate calls prohibit patterns/episodes crossing explicit regime boundaries.
+                for pattern in anomaly_patterns([temporary], ds.id, cusum_k=cfg.cusum_k,
+                        cusum_h=cfg.cusum_h, moving_range_threshold=cfg.moving_range_threshold):
+                    pattern["id"] = f"segment:{segment_number}:{pattern['id']}"
+                    patterns.append(pattern)
+                for observation in episodes([temporary], ds.id):
+                    observation["id"] = f"segment:{segment_number}:{observation['id']}"
+                    observations.append(observation)
+            if any(row["signal_maturity"] == "calibrated" for row in method.evidence):
+                method.status = "completed"
+    except TimeoutError as exc:
+        method.status, method.error = "budget_exceeded", str(exc)
+    except (ArithmeticError, ValueError) as exc:
+        method.status, method.error = "failed", f"{type(exc).__name__}: {exc}"
+    method.diagnostics["detection_readiness"] = dict(
+        status="caution" if method.status == "completed" else "not_assessed",
+        reasons=["Sparse frozen phase medians use only eligible training observations; gaps do not count as measurements.",
+                 "Trend and simultaneous seasonal cycles are not estimated by this sparse profile.",
+                 "Point references may cross gaps; consecutive residual patterns cannot."],
+        assessed_segments=sum(s["status"] == "calibrated" for s in method.diagnostics["segments"]))
+    method.limitations = ["Training and calibration origins remain fixed within each explicit regime; no evaluated episode updates the reference.",
+        "Sparse calibration counts usable residuals, not elapsed grid positions. Phase support and readiness remain separate from business severity.",
+        "No probability, health verdict, abuse accuracy, operational permission or delivery-time performance is established."]
+    resolved = dict(datasets=[ds.model_dump(mode="json")], config=cfg.model_dump(mode="json"),
+                    context=context.model_dump(mode="json"))
+    method.runtime_seconds = time.perf_counter() - started
+    status = "partial" if method.status in ("budget_exceeded", "failed") else method.status
+    return Result(run_id=str(uuid.uuid4()), status=status,
+        input_fingerprint=hashlib.sha256(json.dumps(resolved, sort_keys=True).encode()).hexdigest(),
+        resolved_config=resolved, dependencies=versions(), data_quality=quality, methods=[method],
+        observations=observations, anomaly_patterns=patterns, limitations=method.limitations,
+        suggested_checks=["Inspect coverage, original values, arrival timing and phase support before interpreting departures."],
+        runtime_seconds=time.perf_counter()-started,
+        stop_reason="runtime_budget" if method.status == "budget_exceeded" else "method_failure" if method.status == "failed" else "completed")
+
+
+def active_resets(points, times, zone=None, as_of=None):
+    if as_of:
+        cutoff = timestamp(as_of, zone)
+        points = [p for p in points if (isinstance(p, int) and p < len(times)) or
+                  (isinstance(p, str) and timestamp(p + "T00:00:00Z" if re.fullmatch(r"\d{4}-\d{2}-\d{2}", p) else p, zone) <= cutoff)]
+    return resolve_reset_points(points, times, zone)
+
+
+def safe_median(values):
+    ordered = sorted(values)
+    middle = len(ordered)//2
+    return statistics.mean(ordered[middle-1:middle+1]) if len(ordered)%2 == 0 else ordered[middle]
+
+
+def analyze_contiguous(ds, cfg, context, budget, prepared, resets):
+    """Recipes requiring complete model inputs restart at every unavailable span."""
+    from .recipes import analyze_dataset, _namespace_result
+    times, values, quality = prepared
+    boundaries = sorted(set([0, len(values), *resets] +
+        [position for i, v in enumerate(values) if v is None for position in (i, i+1)]))
+    methods, observations, patterns, spans = [], [], [], []
+    seen_observation = False
+    policy = ds.data_policy or DataPolicy()
+    for start, stop in zip(boundaries, boundaries[1:]):
+        try:
+            budget.check()
+        except TimeoutError as exc:
+            methods.append(MethodResult(id="contiguous_budget", status="budget_exceeded",
+                           applicability="remaining spans were not evaluated", error=str(exc)))
+            break
+        if start == stop or values[start] is None:
+            continue
+        span = dict(start=start, stop=stop, eligible_count=stop-start)
+        spans.append(span)
+        if seen_observation and start not in resets and policy.reference_gaps != "restart":
+            span.update(status="inapplicable", reason="post_gap_learning_requires_explicit_reset_or_restart")
+            continue
+        seen_observation = True
+        if stop-start <= cfg.season_length:
+            continue
+        child = Dataset(id=ds.id, timestamps=times[start:stop], values=values[start:stop],
+                        frequency=ds.frequency, units=ds.units, entity=ds.entity)
+        settings = cfg.model_copy(update={"reset_points": []})
+        result = analyze_dataset(child, settings, context, budget)
+        _namespace_result(result, f"contiguous:{start}")
+        for method in result.methods:
+            method.parameters["source_grid_offset"] = start
+            method.parameters["profile_revision"] = "contiguous-spans-v1"
+            method.parameters["sparse_support"] = "complete contiguous spans only; post-gap learning needs explicit reset/restart"
+            for row in method.evidence:
+                row["index"] += start
+        for item in result.observations:
+            item["triggering_samples"] = [i+start for i in item["triggering_samples"]]
+        for item in result.anomaly_patterns:
+            item["triggering_samples"] = [i+start for i in item["triggering_samples"]]
+            item["detection_index"] += start
+        methods.extend(result.methods)
+        observations.extend(result.observations)
+        patterns.extend(result.anomaly_patterns)
+    if not methods:
+        methods = [MethodResult(id="contiguous_segments", status="insufficient_history",
+                                applicability="no complete span supplies even a seasonal reference")]
+    quality["contiguous_spans"] = spans
+    limited = any(m.status in ("failed", "budget_exceeded") for m in methods)
+    status = "partial" if limited else "completed" if any(m.status == "completed" for m in methods) else "insufficient_history"
+    resolved = dict(datasets=[ds.model_dump(mode="json")], config=cfg.model_dump(mode="json"),
+                    context=context.model_dump(mode="json"))
+    return Result(run_id=str(uuid.uuid4()), status=status,
+        input_fingerprint=hashlib.sha256(json.dumps(resolved, sort_keys=True).encode()).hexdigest(),
+        resolved_config=resolved, dependencies=versions(), data_quality=quality, methods=methods,
+        observations=observations, anomaly_patterns=patterns,
+        limitations=["This recipe uses complete contiguous spans only; no calendar gap is compressed. Post-gap learning requires an explicit reset or reference_gaps=restart.",
+                     "Adaptive spans retain the existing rolling-reference admission policy; an episode quarantine policy is not implemented."],
+        suggested_checks=["Inspect complete-span support and restart origins."],
+        runtime_seconds=sum(m.runtime_seconds for m in methods), stop_reason="runtime_budget" if budget.remaining == 0 else "method_failure" if limited else "completed")
diff --git a/src/anomalyzer/sparse_views.py b/src/anomalyzer/sparse_views.py
new file mode 100644
index 0000000..0b4ddf4
--- /dev/null
+++ b/src/anomalyzer/sparse_views.py
@@ -0,0 +1,206 @@
+"""Schema-1.1 native and complete-day evidence with independent view eligibility."""
+from __future__ import annotations
+
+from datetime import datetime, timedelta, timezone
+import hashlib
+import json
+import math
+import time
+import uuid
+
+from .budget import RuntimeBudget
+from .cases import compose_cases, target_entities
+from .contracts import Context, DataPolicy, DatasetV11, RelationshipResult, RequestV11, ResultV11
+from .registry import versions
+from .sparse import active_resets, cadence_seconds, normalize
+from .validation import timestamp
+
+
+def view_cases(datasets, relationships, request):
+    result = []
+    for resolution in sorted({item.resolution for item in datasets + relationships if item.resolution}):
+        cases = compose_cases([d for d in datasets if d.resolution == resolution],
+                              [r for r in relationships if r.resolution == resolution],
+                              target_entities(request))
+        for case in cases:
+            case.resolution = resolution
+            case.id = f"{resolution}:{case.id}"
+            case.revision_id = f"{resolution}:{case.revision_id}"
+            case.non_claims.append("Native and completed-day views reuse the same observations; they are not independent confirmation.")
+        result.extend(cases)
+    return result
+
+
+def label_view(result, label):
+    """Namespace evidence identifiers without changing original source IDs or times."""
+    for collection in (result.dataset_results, result.relationship_results):
+        for item in collection:
+            item.resolution = label
+            evidence_ids, pattern_ids, method_ids = {}, {}, {}
+            for method in item.methods:
+                old = method.id
+                method.id = f"{label}:{old}"
+                method_ids[old] = method.id
+                for row in method.evidence:
+                    old_id = row["id"]
+                    row["id"] = f"{label}:{old_id}"
+                    row["method"] = method.id
+                    evidence_ids[old_id] = row["id"]
+            for pattern in item.anomaly_patterns:
+                old_id = pattern["id"]
+                pattern["id"] = f"{label}:{old_id}"
+                pattern_ids[old_id] = pattern["id"]
+                pattern["evidence_refs"] = [evidence_ids.get(ref, ref) for ref in pattern["evidence_refs"]]
+            for observation in item.observations:
+                observation["id"] = f"{label}:{observation['id']}"
+                observation["evidence_refs"] = [evidence_ids.get(ref, ref) for ref in observation["evidence_refs"]]
+                observation["supporting_methods"] = [method_ids.get(ref, ref) for ref in observation["supporting_methods"]]
+                observation["quiet_methods"] = [method_ids.get(ref, ref) for ref in observation["quiet_methods"]]
+            for assessment in item.assessments:
+                assessment.id = f"{label}:{assessment.id}"
+                assessment.evidence_refs = [evidence_ids.get(ref, ref) for ref in assessment.evidence_refs]
+                if "method" in assessment.baseline:
+                    assessment.baseline["method"] = method_ids.get(assessment.baseline["method"], assessment.baseline["method"])
+                if "pattern_ids" in assessment.baseline:
+                    assessment.baseline["pattern_ids"] = [pattern_ids.get(ref, ref) for ref in assessment.baseline["pattern_ids"]]
+            item.data_quality["resolution"] = label
+
+
+def daily_dataset(ds, cfg, context, budget):
+    """Sum only wholly observed finalized additive children. Never average ratios."""
+    times, values, quality = normalize(ds, cfg, context, budget)
+    measurements = quality["rule_values"]
+    policy = ds.data_policy or DataPolicy()
+    unsupported = None
+    if policy.observation_basis != "calendar":
+        unsupported = "completed-day aggregation is unavailable on event_index"
+        seconds = None
+    else:
+        seconds = cadence_seconds(ds.frequency)
+        if seconds >= 86400 or 86400 % seconds:
+            unsupported = "source cadence must evenly divide a fixed 24-hour day"
+    if policy.measurement != "additive_volume" or cfg.aggregate_function != "sum":
+        unsupported = "completed-day view requires additive_volume and sum; per-event percentages/ratios are not averaged"
+    if unsupported:
+        return (DatasetV11(id=ds.id, entity=ds.entity, units=ds.units, timestamps=[], values=[],
+                           frequency="1d", data_policy=DataPolicy()),
+                [dict(complete=False, reason=unsupported)], set())
+    anchor = timestamp(cfg.aggregate_anchor, ds.timezone) if cfg.aggregate_anchor else (
+        datetime.fromisoformat(times[0]).replace(hour=0, minute=0, second=0, microsecond=0) if times else None)
+    buckets = {}
+    for index, observed_at in enumerate(times):
+        budget.check()
+        moment = datetime.fromisoformat(observed_at)
+        if seconds and anchor and (moment-anchor).total_seconds() % seconds:
+            raise ValueError("source timestamps are off the aggregate_anchor grid")
+        bucket = int((moment-anchor).total_seconds() // 86400)
+        buckets.setdefault(bucket, []).append(index)
+    daily_times, daily_values, coverage, records, available_at = [], [], [], [], []
+    reset_indexes = active_resets(cfg.reset_points, times, ds.timezone, context.as_of)
+    reset_times = [datetime.fromisoformat(times[index]) for index in reset_indexes]
+    daily_reset_times = set()
+    if buckets:
+        for bucket in range(min(buckets), max(buckets)+1):
+            budget.check()
+            start = anchor + timedelta(days=bucket)
+            indexes = buckets.get(bucket, [])
+            expected = 86400 // seconds if seconds and 86400 % seconds == 0 else None
+            inside_resets = [t for t in reset_times if start <= t < start+timedelta(days=1)]
+            for t in inside_resets:
+                daily_reset_times.add((start if t == start else start+timedelta(days=1)).isoformat())
+            complete = (not unsupported and len(indexes) == expected and
+                        all(measurements[i] is not None for i in indexes))
+            excluded = any(t != start for t in inside_resets) or any(quality["samples"][i]["coverage"] == "excluded" for i in indexes)
+            value = math.fsum(measurements[i] for i in indexes) if complete else None
+            if value is not None and not math.isfinite(value):
+                value, complete = None, False
+            reason = ("missing_or_unfinished_children" if not complete else
+                      "reset_inside_aggregate" if any(t != start for t in inside_resets) else
+                      "excluded_model_child" if excluded else None)
+            daily_times.append(start.isoformat())
+            daily_values.append(value)
+            known_arrivals = [quality["samples"][i]["available_at"] for i in indexes if quality["samples"][i]["available_at"] and measurements[i] is not None]
+            available_at.append(max([*known_arrivals, *[times[i] for i in indexes]]) if known_arrivals else None)
+            coverage.append("excluded" if complete and excluded else "complete" if complete else "incomplete")
+            records.append(dict(timestamp=start.isoformat(), complete=bool(complete), reason=reason,
+                expected_children=expected, usable_children=sum(measurements[i] is not None for i in indexes),
+                model_eligible=bool(complete and not excluded),
+                policy_zero_children=sum(quality["samples"][i]["reason"] == "confirmed_empty_additive_volume" for i in indexes),
+                source_input_indexes=[raw for i in indexes for raw in quality["samples"][i]["source_indexes"]],
+                source_grid_indexes=indexes, observed_sum=(math.fsum(measurements[i] for i in indexes if measurements[i] is not None)
+                    if any(measurements[i] is not None for i in indexes) else None),
+                aggregate_value=value))
+    child = DatasetV11(id=ds.id, entity=ds.entity, units=ds.units,
+        timestamps=daily_times, values=daily_values, frequency="1d", coverage=coverage, available_at=available_at,
+        data_policy=DataPolicy(measurement="additive_volume", reference_gaps=policy.reference_gaps,
+            minimum_phase_samples=policy.minimum_phase_samples, change_max_gap_seconds=policy.change_max_gap_seconds),
+        rules=ds.completed_period_rules)
+    return child, records, daily_reset_times
+
+
+def analyze_multi_request(request, budget=None):
+    from .orchestrator import analyze_relationships
+    started = time.perf_counter()
+    budget = budget or RuntimeBudget(request.config.max_runtime_seconds)
+    # Every child source is explicitly prepared so incomplete/unknown grid slots survive.
+    native_sources = [d.model_copy(update={"data_policy": d.data_policy or DataPolicy()}) for d in request.datasets]
+    native_config = request.config.model_copy(update={"recipe": "seasonal-residual-v1",
+        "season_length": request.config.intraday_season_length or request.config.season_length})
+    native_request = request.model_copy(update={"datasets": native_sources, "config": native_config})
+    native = analyze_relationships(native_request, _budget=budget)
+    label_view(native, "intraday")
+    sources, aggregation, reset_times = [], {}, set()
+    for ds in native_sources:
+        try:
+            child, records, resets = daily_dataset(ds, request.config, request.context, budget)
+            sources.append(child)
+            aggregation[ds.id] = records
+            reset_times.update(resets)
+        except (ValueError, TimeoutError) as exc:
+            sources.append(DatasetV11(id=ds.id, entity=ds.entity, units=ds.units,
+                                     timestamps=[], values=[], frequency="1d", data_policy=DataPolicy()))
+            aggregation[ds.id] = {"error": str(exc)}
+    all_times = {t for ds in sources for t in ds.timestamps}
+    daily_config = request.config.model_copy(update={"recipe": "seasonal-residual-v1",
+        "season_length": request.config.completed_period_season_length,
+        "reset_points": sorted(t for t in reset_times if t in all_times and any(t != ds.timestamps[0] for ds in sources if ds.timestamps))})
+    daily_relationships = [r.model_copy(update={"rules": r.completed_period_rules})
+                           for r in request.relationships if r.kind in ("ratio", "difference")]
+    daily_request = request.model_copy(update={"datasets": sources, "config": daily_config, "relationships": daily_relationships})
+    daily = analyze_relationships(daily_request, _budget=budget)
+    for relationship in request.relationships:
+        if relationship.kind not in ("ratio", "difference"):
+            daily.relationship_results.append(RelationshipResult(relationship_id=relationship.id, kind=relationship.kind,
+                status="inapplicable", definition=relationship.model_dump(mode="json"),
+                applicability="completed-day view supports ratio-of-sums and differences of compatible additive sums only",
+                data_quality={"unavailable_count": 0}, lineage=[], methods=[], observations=[], anomaly_patterns=[], assessments=[],
+                limitations=["No hourly lag, normalization scale or joint-condition threshold was silently reinterpreted as daily input."]))
+    label_view(daily, "completed_period")
+    for item in daily.dataset_results:
+        item.data_quality["aggregation"] = aggregation[item.dataset_id]
+        item.limitations.append("Incomplete sums are progress only, never completed consumption; ratio views divide summed compatible source quantities.")
+    for relationship in daily.relationship_results:
+        for point in relationship.lineage:
+            for source_id, source_index in point.source_indexes.items():
+                records = aggregation[source_id]
+                if isinstance(records, list) and 0 <= source_index < len(records):
+                    point.source_input_indexes[source_id] = records[source_index].get("source_input_indexes", [])
+    datasets = native.dataset_results + daily.dataset_results
+    relationships = native.relationship_results + daily.relationship_results
+    status = "completed" if native.status == daily.status == "completed" else "partial"
+    resolved = request.model_dump(mode="json")
+    return ResultV11(run_id=str(uuid.uuid4()), status=status,
+        input_fingerprint=hashlib.sha256(json.dumps(resolved, sort_keys=True).encode()).hexdigest(),
+        resolved_config=resolved, dependencies=versions(), dataset_results=datasets,
+        relationship_results=relationships, cases=view_cases(datasets, relationships, request),
+        limitations=["Views are correlated evidence, not independent incident votes.",
+                     "Schema-1.1 multi-resolution uses a frozen seasonal child recipe and fixed UTC 24-hour periods; adaptive children and local-calendar/DST aggregation remain unsupported.",
+                     "Caller rules are evaluated in each view's declared units; change and sum thresholds require a suitable per-view interpretation."],
+        runtime_seconds=time.perf_counter()-started, stop_reason="runtime_budget" if budget.remaining == 0 else "completed")
+
+
+def analyze_sparse_views(ds, cfg, context, budget=None):
+    if budget:
+        budget.check()
+    return analyze_multi_request(RequestV11(schema_version="1.1", datasets=[ds], config=cfg,
+                                 context=Context.model_validate(context or {})), budget)
diff --git a/tests/test_mcp.py b/tests/test_mcp.py
index baa8b7a..fe21695 100644
--- a/tests/test_mcp.py
+++ b/tests/test_mcp.py
@@ -338,7 +338,7 @@ def test_schema_11_episode_is_paged_and_has_relationship_evidence():
         "orders": 14, "qualified_visits": 14}
 
 
-def test_inapplicable_method_keeps_descriptive_candidate_visible():
+def test_sparse_insufficient_history_keeps_descriptive_candidate_visible():
     request = {
         "schema_version": "1.1",
         "datasets": [{"id": "series", "entity": {"account": "example"},
@@ -350,7 +350,7 @@ def test_inapplicable_method_keeps_descriptive_candidate_visible():
     summary = handle({"jsonrpc": "2.0", "id": 1, "method": "tools/call",
                       "params": {"name": "analyze", "arguments": {"request": request}}}
                      )["result"]["structuredContent"]
-    assert summary["status"] == "inapplicable"
+    assert summary["status"] == "insufficient_history"
     candidate = next(item for item in summary["findings_preview"]
                      if item["kind"] == "candidate")
     detail = handle({"jsonrpc": "2.0", "id": 2, "method": "tools/call",
diff --git a/tests/test_release.py b/tests/test_release.py
index 827657d..6adfb41 100644
--- a/tests/test_release.py
+++ b/tests/test_release.py
@@ -78,7 +78,7 @@ def test_ci_stages_draft_without_admin_api_or_publication(monkeypatch, tmp_path)
 def test_changed_artifact_fails_checksum_verification(monkeypatch, tmp_path):
     (tmp_path / "wheel.whl").write_bytes(b"changed after testing")
     (tmp_path / "release-manifest.json").write_text(json.dumps({
-        "version": release.version(), "source_revision": "a" * 40, "tag": "v0.1.0",
+        "version": release.version(), "source_revision": "a" * 40, "tag": f"v{release.version()}",
         "artifacts": {"wheel.whl": {"sha256": "0" * 64, "size": 21}}}))
     with pytest.raises(AssertionError, match="wheel.whl"):
-        release.verify(tmp_path, "a" * 40, "v0.1.0")
+        release.verify(tmp_path, "a" * 40, f"v{release.version()}")
diff --git a/tests/test_sparse.py b/tests/test_sparse.py
new file mode 100644
index 0000000..ce1f2bf
--- /dev/null
+++ b/tests/test_sparse.py
@@ -0,0 +1,469 @@
+"""Sparse source semantics, causal replay, bounded preparation and public interfaces."""
+import copy
+from datetime import datetime, timedelta, timezone
+import json
+import subprocess
+import sys
+
+import pytest
+
+from anomalyzer import analyze, analyze_series, analyze_relationships, get_case, replay_policy
+from anomalyzer.contracts import DataPolicy, DatasetV11, Request, RequestV11, ResultV11
+from anomalyzer.mcp import handle
+
+
+def body(values=None, **dataset):
+    values = [10, 11, 9, 10, 9, 10, 11, 10, None, 10, 11, 9, 10, 30] if values is None else values
+    times = [(datetime(2026, 1, 1, tzinfo=timezone.utc)+timedelta(hours=i)).isoformat() for i in range(len(values))]
+    return {"schema_version": "1.1", "datasets": [{"id": "series", "entity": {"customer": "test"},
+        "timestamps": times, "values": values, "frequency": "1h", **dataset}],
+        "config": {"season_length": 1, "training_size": 4, "calibration_size": 3, "trend": "none"}}
+
+
+def source(request):
+    return analyze(request).dataset_results[0]
+
+
+def evidence(item):
+    return [row for method in item.methods for row in method.evidence]
+
+
+def test_omitted_positions_equal_explicit_null_without_compressing_time():
+    explicit = body()
+    omitted = copy.deepcopy(explicit)
+    del omitted["datasets"][0]["timestamps"][8]
+    del omitted["datasets"][0]["values"][8]
+    left, right = source(explicit), source(omitted)
+    assert evidence(left) == evidence(right)
+    assert left.data_quality["grid_count"] == right.data_quality["grid_count"] == 14
+    assert left.data_quality["observation_count"] == right.data_quality["observation_count"] == 13
+    assert left.data_quality["zero_filled_count"] == right.data_quality["zero_filled_count"] == 0
+    assert right.data_quality["samples"][8]["source_index"] == -1
+    assert not any(row["index"] == 8 for row in evidence(right))
+
+
+def test_zero_is_only_authorized_for_complete_empty_additive_volume():
+    request = body([2, None, None, 0, 9], data_policy={"missing": "zero_for_empty", "measurement": "additive_volume"},
+                   coverage=["complete", "empty", "unknown", "complete", "incomplete"])
+    result = source(request)
+    rows = result.data_quality["samples"]
+    assert [r["value"] for r in rows] == [2, 0, None, 0, None]
+    assert [r["original_value"] for r in rows] == [2, None, None, 0, 9]
+    assert rows[1]["reason"] == "confirmed_empty_additive_volume"
+    assert rows[3]["reason"] is None
+    assert result.data_quality["zero_filled_count"] == 1
+    assert result.data_quality["latest_evidence_state"] == "insufficient_evidence"
+
+
+def test_global_collection_contract_authorizes_omitted_slots_but_not_explicit_unknown():
+    request = body([1, None, 2], data_policy={"missing": "zero_for_empty", "measurement": "additive_volume", "absence_means_zero": True})
+    request["datasets"][0]["timestamps"][2] = "2026-01-01T03:00:00Z"
+    assert [r["value"] for r in source(request).data_quality["samples"]] == [1, 0, 0, 2]
+    request["datasets"][0]["coverage"] = ["complete", "unknown", "complete"]
+    assert [r["value"] for r in source(request).data_quality["samples"]] == [1, None, 0, 2]
+
+
+@pytest.mark.parametrize("policy", [
+    {"missing": "zero_for_empty"}, {"absence_means_zero": True},
+    {"observation_basis": "event_index", "grid_start": "2026-01-01T00:00:00Z"},
+])
+def test_incompatible_policy_is_rejected(policy):
+    with pytest.raises(ValueError):
+        DataPolicy.model_validate(policy)
+
+
+@pytest.mark.parametrize("mutation,reason", [
+    (lambda d: d["timestamps"].__setitem__(2, "2026-01-01T02:30:00Z"), "off"),
+    (lambda d: d["data_policy"].update(grid_end="2036-01-01T00:00:00Z"), "max_points"),
+    (lambda d: d["coverage"].__setitem__(1, "empty"), "nonzero"),
+])
+def test_invalid_or_unbounded_grid_fails_visibly(mutation, reason):
+    request = body([1, 2, 3], data_policy={}, coverage=["complete"]*3)
+    mutation(request["datasets"][0])
+    result = source(request)
+    assert result.status == "failed" and reason in result.error
+
+
+@pytest.mark.parametrize("policy,value", [("sum", 5), ("mean", 2.5)])
+def test_declared_duplicate_aggregation_preserves_provenance(policy, value):
+    request = body([2, 3, 4], data_policy={})
+    request["datasets"][0]["timestamps"][1] = request["datasets"][0]["timestamps"][0]
+    request["config"]["duplicate_policy"] = policy
+    result = source(request)
+    assert result.status != "failed"
+    assert result.data_quality["samples"][0]["value"] == value
+    assert result.data_quality["duplicate_sources"][0]["original_values"] == [2, 3]
+
+
+def test_expansion_uses_shared_point_budget_and_keeps_valid_sibling():
+    request = body([1, 2], data_policy={})
+    request["datasets"][0]["timestamps"][-1] = "2026-01-01T03:00:00Z"
+    request["datasets"].append({**copy.deepcopy(request["datasets"][0]), "id": "second"})
+    request["config"]["max_points"] = 5
+    result = analyze(request)
+    assert result.status == "partial"
+    assert result.dataset_results[0].status != "failed"
+    assert "shared max_points" in result.dataset_results[1].error
+
+
+def test_sparse_phase_references_match_elapsed_weekday_and_never_borrow_evaluation():
+    request = body([100 + [0, 10, 20, 30, 40, 50, 60][i % 7] for i in range(60)], data_policy={})
+    request["datasets"][0]["values"][8] = None
+    request["config"].update(season_length=7, training_size=21, calibration_size=3)
+    original = source(request)
+    refs = original.methods[0].diagnostics["segments"][0]["phase_references"]
+    for ref in refs:
+        assert all(i % 7 == ref["phase"] for i in ref["indexes"])
+    changed = copy.deepcopy(request)
+    changed["datasets"][0]["values"][-10:] = [999]*10
+    later = source(changed)
+    assert later.methods[0].diagnostics["segments"][0]["phase_references"] == refs
+    assert [(r["index"], r["expected"]) for r in evidence(original)] == [(r["index"], r["expected"]) for r in evidence(later)]
+
+
+def test_sparse_phase_without_support_is_explicitly_unavailable():
+    request = body([10]*30, data_policy={})
+    request["config"].update(season_length=24, training_size=4)
+    result = source(request)
+    assert result.status == "insufficient_history"
+    assert any(row["reason"] == "no_training_support_for_calendar_phase" for row in result.methods[0].diagnostics["unavailable"])
+
+
+def test_gap_and_reset_do_not_join_nelson_or_cusum_streams():
+    request = body([100]*4 + [99, 100, 101] + [102]*4 + [None] + [102]*7, data_policy={})
+    request["config"]["point_threshold"] = 100
+    item = source(request)
+    assert not any(p["rule"] == "nelson_rule_2" for p in item.anomaly_patterns)
+    for pattern in item.anomaly_patterns:
+        assert not (min(pattern["triggering_samples"]) < 11 < max(pattern["triggering_samples"]))
+    request["config"]["reset_points"] = [14]
+    reset = source(request)
+    for pattern in reset.anomaly_patterns:
+        assert not (min(pattern["triggering_samples"]) < 14 <= max(pattern["triggering_samples"]))
+
+
+def test_first_range_rule_survives_missing_history_and_policy_replay():
+    request = body([100], data_policy={}, rules=[{"id": "measured-limit", "kind": "acceptable_range", "minimum": 0, "maximum": 20}])
+    result = analyze(request)
+    assert result.status == "insufficient_history"
+    assert result.dataset_results[0].assessments[0].criterion_met
+    assert not result.cases[0].action_eligible
+    replayed = replay_policy(result, {"allow_explicit_rules": True})
+    assert get_case(replayed, replayed.cases[0].id).action_eligible
+    assert not result.cases[0].action_eligible
+
+
+def test_changes_require_comparable_prior_and_reset_early_population():
+    request = body([0, 20, None, 100, 120, 900], data_policy={}, rules=[
+        {"id": "step", "kind": "maximum_absolute_change", "threshold": 30},
+        {"id": "relative", "kind": "maximum_relative_change", "threshold": 10}])
+    request["config"]["reset_points"] = [5]
+    result = source(request)
+    assert not any(a.criterion_met for a in result.assessments)
+    assert result.assessments[-1].comparison_samples == 0
+    assert result.assessments[1].baseline["unavailable_rules"][0]["reason"] == "relative_change_from_zero"
+
+
+def test_excluded_model_value_is_not_training_but_can_violate_measured_rule():
+    request = body([10, 999, 10, 10, 10, 9, 10, 11, 20], data_policy={},
+        coverage=["complete", "excluded"]+["complete"]*7,
+        rules=[{"id": "range", "kind": "acceptable_range", "minimum": 0, "maximum": 500}])
+    result = source(request)
+    assert result.assessments[1].criterion_met
+    assert 1 not in result.methods[0].diagnostics["segments"][0]["training_indexes"]
+
+
+def test_event_index_keeps_real_times_and_optional_time_separation():
+    request = body([10, 100], data_policy={"observation_basis": "event_index", "change_max_gap_seconds": 3600},
+        rules=[{"id": "step", "kind": "maximum_absolute_change", "threshold": 20}])
+    ds = request["datasets"][0]
+    ds.pop("frequency")
+    ds["timestamps"][1] = "2026-02-01T00:00:00Z"
+    result = analyze_series(ds, request["config"])
+    assert isinstance(result, ResultV11)
+    assert result.dataset_results[0].data_quality["coordinate"] == "event_index"
+    assert result.dataset_results[0].assessments[-1].timestamp == "2026-02-01T00:00:00+00:00"
+    assert not result.dataset_results[0].assessments[-1].criterion_met
+    request["config"]["season_length"] = 7
+    assert source(request).methods[0].status == "inapplicable"
+
+
+def test_as_of_uses_original_revision_until_correction_was_available():
+    request = body([10]*12, data_policy={})
+    ds = request["datasets"][0]
+    ds["revisions"] = [{"timestamp": ds["timestamps"][8], "value": 100,
+        "available_at": "2026-01-02T00:00:00Z", "revision_id": "corrected"}]
+    request["context"] = {"as_of": ds["timestamps"][-1]}
+    early = source(request)
+    no_revision = copy.deepcopy(request)
+    no_revision["datasets"][0]["revisions"] = []
+    assert evidence(early) == evidence(source(no_revision))
+    request["context"]["as_of"] = "2026-01-02T00:00:00Z"
+    later = source(request)
+    assert later.data_quality["samples"][8]["value"] == 100
+    assert later.data_quality["samples"][8]["revision_id"] == "corrected"
+
+
+def test_late_value_is_unknown_before_arrival_and_cannot_train():
+    request = body([10]*12, data_policy={})
+    request["datasets"][0]["available_at"] = [None]*12
+    request["datasets"][0]["available_at"][1] = "2026-01-02T00:00:00Z"
+    request["context"] = {"as_of": "2026-01-01T11:00:00Z"}
+    result = source(request)
+    assert result.data_quality["samples"][1]["value"] is None
+    assert result.data_quality["samples"][1]["original_value"] is None
+    assert result.data_quality["samples"][1]["reason"] == "not_available_as_of"
+
+
+def test_future_as_of_and_reset_do_not_change_previous_evidence():
+    request = body([10, 11, 9, 10, 9, 10, 11, None, 10, 10, 30, 10, 11, 12], data_policy={})
+    request["config"]["reset_points"] = [12]
+    full = source(request)
+    request["context"] = {"as_of": request["datasets"][0]["timestamps"][10]}
+    prefix = source(request)
+    assert evidence(prefix) == [r for r in evidence(full) if r["index"] <= 10]
+    assert [a.model_dump() for a in prefix.assessments] == [a.model_dump() for a in full.assessments if a.index <= 10]
+
+
+def ratio_request(top, bottom):
+    request = body(top, data_policy={"measurement": "additive_volume"}, units="dollars")
+    request["datasets"].append({**copy.deepcopy(request["datasets"][0]), "id": "calls", "units": "calls", "values": bottom})
+    request["relationships"] = [{"id": "unit-cost", "kind": "ratio", "numerator": "series", "denominator": "calls", "units": "dollars_per_call"}]
+    return request
+
+
+def test_relationship_missing_zero_and_tiny_denominator_keep_lineage():
+    request = ratio_request([1, None, 1, 1e308], [0, 10, None, 1e-308])
+    relationship = analyze(request).relationship_results[0]
+    assert [row.value for row in relationship.lineage] == [None]*4
+    assert [row.unavailable_reason for row in relationship.lineage] == ["zero_denominator", "missing_source_value", "missing_source_value", "nonfinite_derived_value"]
+    assert relationship.lineage[0].source_values == {"series": 1, "calls": 0}
+    assert relationship.lineage[1].source_coverage["series"] == "unknown"
+
+
+def test_complete_daily_ratio_divides_sums_and_missing_child_invalidates_only_its_view():
+    request = ratio_request([1, 9]*24, [100, 10]*24)
+    request["config"]["recipe"] = "multi-resolution-v1"
+    result = analyze(request)
+    daily = next(r for r in result.relationship_results if r.resolution == "completed_period")
+    assert [r.value for r in daily.lineage] == pytest.approx([10/110, 10/110])
+    request["datasets"][0]["values"][5] = None
+    result = analyze(request)
+    item = next(d for d in result.dataset_results if d.dataset_id == "series" and d.resolution == "completed_period")
+    assert item.data_quality["aggregation"][0]["aggregate_value"] is None
+    assert item.data_quality["aggregation"][0]["usable_children"] == 23
+    assert any(d.resolution == "intraday" and d.status != "failed" for d in result.dataset_results)
+    replayed = replay_policy(result, {"allow_explicit_rules": True})
+    ResultV11.model_validate_json(replayed.model_dump_json())
+
+
+def test_daily_rules_are_not_silently_reused_from_hourly_rules():
+    request = body([1]*48, data_policy={"measurement": "additive_volume"},
+        rules=[{"id": "hourly", "kind": "acceptable_range", "minimum": 0, "maximum": 2}])
+    request["config"]["recipe"] = "multi-resolution-v1"
+    result = analyze(request)
+    daily = next(d for d in result.dataset_results if d.resolution == "completed_period")
+    assert not any(a.criterion_met for a in daily.assessments)
+
+
+def test_adaptive_sparse_support_uses_complete_spans_with_global_indexes():
+    request = body([10]*12+[None]+[10]*12+[50], data_policy={"reference_gaps":"restart"})
+    request["config"].update(recipe="adaptive-seasonal-v1", trend="auto")
+    result = source(request)
+    assert result.status == "completed"
+    assert any(m.parameters.get("source_grid_offset") == 13 for m in result.methods)
+    assert max(r["index"] for r in evidence(result)) == 25
+    assert not any(r["index"] == 12 for r in evidence(result))
+
+
+def test_adaptive_gap_does_not_silently_train_a_post_gap_episode():
+    request = body([10]*12+[None]+[100]*20, data_policy={})
+    request["config"].update(recipe="adaptive-seasonal-v1", trend="auto")
+    result = source(request)
+    assert all(row["index"] < 12 for row in evidence(result))
+    assert result.data_quality["contiguous_spans"][-1]["reason"] == "post_gap_learning_requires_explicit_reset_or_restart"
+
+
+def test_legacy_schema_still_rejects_new_policy_and_keeps_missing_inapplicable():
+    request = body()
+    request.pop("schema_version")
+    assert analyze(request).status == "inapplicable"
+    request["datasets"][0]["data_policy"] = {}
+    with pytest.raises(ValueError):
+        Request.model_validate(request)
+
+
+def stable(value):
+    if isinstance(value, dict):
+        return {k: stable(v) for k, v in value.items() if k not in ("run_id", "runtime_seconds")}
+    return [stable(v) for v in value] if isinstance(value, list) else value
+
+
+def test_sparse_python_cli_and_mcp_full_result_parity():
+    request = body(data_policy={})
+    direct = analyze_relationships(request).model_dump(mode="json")
+    cli = subprocess.run([sys.executable, "-m", "anomalyzer", "analyze", "-", "--input-format", "json", "--format", "json"],
+                         input=json.dumps(request), capture_output=True, text=True, timeout=30)
+    assert cli.returncode == 0, cli.stderr
+    mcp = handle({"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "analyze_relationships", "arguments": request}})
+    assert stable(direct) == stable(json.loads(cli.stdout)) == stable(mcp["result"]["structuredContent"])
+
+
+def test_internal_incomplete_status_breaks_models_and_changes_without_losing_grid():
+    request = body([10]*10+[100]+[10]*10, period_statuses=["complete"]*10+["incomplete"]+["complete"]*10)
+    result = source(request)
+    assert result.data_quality["samples"][10]["original_value"] == 100
+    assert result.data_quality["samples"][10]["original_period_status"] == "incomplete"
+    assert not any(a.index == 10 for a in result.assessments)
+    assert next(a for a in result.assessments if a.index == 11).comparison_samples == 0
+    assert not any(r["index"] == 10 for r in evidence(result))
+
+
+def test_restart_policy_cannot_reuse_a_prior_regime_reference():
+    request = body([10]*10+[None]+[100], data_policy={})
+    reuse = source(request)
+    assert any(r["index"] == 11 and r["triggers"] for r in evidence(reuse))
+    request["datasets"][0]["data_policy"]["reference_gaps"] = "restart"
+    restart = source(request)
+    assert not any(r["index"] == 11 for r in evidence(restart))
+    assert restart.methods[0].diagnostics["segments"][-1]["training_indexes"] == [11]
+
+
+def test_case_retains_delayed_correction_availability_cutoff():
+    request = body([10], data_policy={}, rules=[{"id":"range","kind":"acceptable_range","minimum":0,"maximum":20}])
+    ds = request["datasets"][0]
+    ds["revisions"] = [{"timestamp":ds["timestamps"][0],"value":100,
+                         "available_at":"2026-01-02T00:00:00Z","revision_id":"r1"}]
+    request["context"] = {"as_of":"2026-01-02T00:00:00Z"}
+    result = analyze(request)
+    assert result.cases[0].event_time == "2026-01-01T00:00:00+00:00"
+    assert result.cases[0].source_availability_cutoff == "2026-01-02T00:00:00+00:00"
+    request["context"]["as_of"] = "2026-01-01T00:00:00Z"
+    assert not analyze(request).cases
+
+
+def test_future_late_event_cannot_move_an_earlier_event_index_or_calendar_origin():
+    for basis in ("calendar", "event_index"):
+        request = body([10]*12, data_policy={"observation_basis":basis})
+        ds = request["datasets"][0]
+        if basis == "event_index":
+            ds.pop("frequency")
+        request["context"] = {"as_of":"2026-01-01T11:00:00Z"}
+        original = source(request)
+        ds["timestamps"].insert(0,"2025-12-31T23:00:00Z")
+        ds["values"].insert(0,999)
+        ds["available_at"] = ["2026-01-02T00:00:00Z"]+[None]*12
+        later = source(request)
+        assert evidence(original) == evidence(later)
+        assert original.data_quality["window"] == later.data_quality["window"]
+
+
+def test_model_exclusion_retains_exact_relationship_range_rules():
+    request = ratio_request([100, 1, 1, 1, 1], [1]*5)
+    request["datasets"][0]["coverage"] = ["excluded"]+["complete"]*4
+    request["relationships"][0]["rules"] = [{"id":"range","kind":"acceptable_range","minimum":0,"maximum":10}]
+    result = analyze(request).relationship_results[0]
+    assert result.lineage[0].value == 100
+    assert result.assessments[0].criterion_met
+    assert not any(row["index"] == 0 for method in result.methods for row in method.evidence)
+
+
+@pytest.mark.parametrize("values", [[0]*20, [1e-308]*20, [1e308, -1e308]])
+def test_degenerate_finite_measurements_keep_valid_rule_results(values):
+    request = body(values, data_policy={}, rules=[{"id":"range","kind":"acceptable_range","minimum":0,"maximum":1}])
+    result = analyze(request)
+    assert result.dataset_results[0].status != "failed"
+    ResultV11.model_validate_json(result.model_dump_json())
+
+
+def test_sibling_failure_and_exhausted_budget_are_visible():
+    request = body([1, 2, 3], data_policy={})
+    request["datasets"].append({**copy.deepcopy(request["datasets"][0]), "id":"bad"})
+    request["datasets"][1]["timestamps"][1] = "2026-01-01T01:30:00Z"
+    result = analyze(request)
+    assert result.status == "partial"
+    assert next(d for d in result.dataset_results if d.dataset_id == "series").status != "failed"
+    request["config"]["max_runtime_seconds"] = 1e-9
+    result = analyze(request)
+    assert result.stop_reason == "runtime_budget"
+    assert all(d.error and "budget" in d.error for d in result.dataset_results)
+
+
+def test_multi_resolution_mcp_drilldown_selects_correct_view_and_replays_case_ids():
+    from anomalyzer.mcp_views import build_findings, get_evidence
+    request = body([1]*60, data_policy={"measurement":"additive_volume"},
+        completed_period_rules=[{"id":"daily","kind":"acceptable_range","minimum":0,"maximum":20}])
+    request["config"]["recipe"] = "multi-resolution-v1"
+    result = analyze(request)
+    assert result.cases and all(c.resolution == "completed_period" for c in result.cases)
+    replayed = replay_policy(result, {"allow_explicit_rules":True})
+    assert {c.id for c in replayed.cases} == {c.id for c in result.cases}
+    assert all(c.action_eligible for c in replayed.cases)
+    doc = result.model_dump(mode="json")
+    findings = build_findings(doc)
+    assert len({f["id"] for f in findings}) == len(findings)
+    for finding in findings:
+        detail = get_evidence(doc, finding["id"])
+        assert detail["finding"]["resolution"] == finding["resolution"]
+
+
+def test_unsupported_daily_relationship_is_explicit_and_does_not_reinterpret_lag():
+    request = ratio_request([1]*48, [1]*48)
+    request["config"]["recipe"] = "multi-resolution-v1"
+    request["relationships"] = [{"id":"lag","kind":"lagged_response","response":"series","predictor":"calls","lag":2,"coefficient":1,"units":"dollars"}]
+    result = analyze(request)
+    daily = next(r for r in result.relationship_results if r.resolution == "completed_period")
+    assert daily.status == "inapplicable" and "lag" in daily.limitations[0]
+
+
+@pytest.mark.parametrize("exclusion", ["reset", "excluded_child"])
+def test_complete_daily_measurement_survives_model_exclusion(exclusion):
+    request = body([1]*48, data_policy={"measurement":"additive_volume"},
+        completed_period_rules=[{"id":"daily","kind":"acceptable_range","minimum":0,"maximum":20}])
+    request["config"]["recipe"] = "multi-resolution-v1"
+    if exclusion == "reset":
+        request["config"]["reset_points"] = [12]
+    else:
+        request["datasets"][0]["coverage"] = ["excluded"]+["complete"]*47
+    result = analyze(request)
+    daily = next(d for d in result.dataset_results if d.resolution == "completed_period")
+    row = daily.data_quality["aggregation"][0]
+    assert row["complete"] and row["aggregate_value"] == 24 and not row["model_eligible"]
+    assert daily.assessments[0].criterion_met
+
+
+def test_duplicate_relationship_lineage_lists_all_original_inputs():
+    request = ratio_request([1, 2, 3], [10, 20, 30])
+    for ds in request["datasets"]:
+        ds["timestamps"][1] = ds["timestamps"][0]
+    request["config"]["duplicate_policy"] = "sum"
+    row = analyze(request).relationship_results[0].lineage[0]
+    assert row.value == 0.1
+    assert row.source_indexes == {"series":-1, "calls":-1}
+    assert row.source_input_indexes == {"series":[0,1], "calls":[0,1]}
+
+
+def test_event_timestamp_ties_are_rejected_without_fabricated_offsets():
+    request = body([1, 2], data_policy={"observation_basis":"event_index"})
+    request["datasets"][0].pop("frequency")
+    request["datasets"][0]["timestamps"][1] = request["datasets"][0]["timestamps"][0]
+    result = source(request)
+    assert result.status == "failed" and "unique event timestamps" in result.error
+
+
+def test_future_arrival_inside_old_unknown_slot_does_not_rewrite_earlier_assessments():
+    request = body(data_policy={})
+    request["context"] = {"as_of":request["datasets"][0]["timestamps"][-1]}
+    original = source(request)
+    request["datasets"][0]["values"][8] = 999
+    request["datasets"][0]["available_at"] = [None]*14
+    request["datasets"][0]["available_at"][8] = "2026-01-02T00:00:00Z"
+    future = source(request)
+    assert evidence(original) == evidence(future)
+    assert [a.model_dump() for a in original.assessments] == [a.model_dump() for a in future.assessments]
+
+
+def test_multi_resolution_recomputation_keeps_profile_parameters_deterministic():
+    request = ratio_request([1,9]*24, [100,10]*24)
+    request["config"]["recipe"] = "multi-resolution-v1"
+    assert stable(analyze(request).model_dump(mode="json")) == stable(analyze(request).model_dump(mode="json"))
diff --git a/uv.lock b/uv.lock
index 61c4db2..da9599e 100644
--- a/uv.lock
+++ b/uv.lock
@@ -30,16 +30,20 @@ dev = [
     { name = "jsonschema" },
     { name = "pytest" },
 ]
+docs = [
+    { name = "markdown-it-py" },
+]
 
 [package.metadata]
 requires-dist = [
     { name = "jsonschema", marker = "extra == 'dev'", specifier = ">=4,<5" },
+    { name = "markdown-it-py", marker = "extra == 'docs'", specifier = "==3.0.0" },
     { name = "pydantic", specifier = ">=2.10,<3" },
     { name = "pytest", marker = "extra == 'dev'", specifier = ">=8,<10" },
     { name = "typing-extensions", specifier = ">=4.12.2" },
     { name = "tzdata", marker = "sys_platform == 'win32'", specifier = ">=2024.1" },
 ]
-provides-extras = ["dev"]
+provides-extras = ["dev", "docs"]
 
 [[package]]
 name = "attrs"
@@ -95,6 +99,27 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/41/45/1a4ed80516f02155c51f51e8cedb3c1902296743db0bbc66608a0db2814f/jsonschema_specifications-2025.9.1-py3-none-any.whl", hash = "sha256:98802fee3a11ee76ecaca44429fda8a41bff98b00a0f2838151b113f210cc6fe", size = 18437 },
 ]
 
+[[package]]
+name = "markdown-it-py"
+version = "3.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "mdurl" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/38/71/3b932df36c1a044d397a1f92d1cf91ee0a503d91e470cbd670aa66b07ed0/markdown-it-py-3.0.0.tar.gz", hash = "sha256:e3f60a94fa066dc52ec76661e37c851cb232d92f9886b15cb560aaada2df8feb", size = 74596 }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/42/d7/1ec15b46af6af88f19b8e5ffea08fa375d433c998b8a7639e76935c14f1f/markdown_it_py-3.0.0-py3-none-any.whl", hash = "sha256:355216845c60bd96232cd8d8c40e8f9765cc86f46880e43a8fd22dc1a1a8cab1", size = 87528 },
+]
+
+[[package]]
+name = "mdurl"
+version = "0.1.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/d6/54/cfe61301667036ec958cb99bd3efefba235e65cdeb9c84d24a8293ba1d90/mdurl-0.1.2.tar.gz", hash = "sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba", size = 8729 }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b3/38/89ba8ad64ae25be8de66a6d463314cf1eb366222074cfda9ee839c56a4b4/mdurl-0.1.2-py3-none-any.whl", hash = "sha256:84008a41e51615a49fc9966191ff91509e3c40b939176e643fd50a5c2196b8f8", size = 9979 },
+]
+
 [[package]]
 name = "packaging"
 version = "26.3"