chore(packs): enroll ci-wall-clock and repin compound packs to 8d07118079

[systemfsoftware-maker]

Summary

Reviews in this repository now load the org-wide ci-wall-clock pack, and the four packs already enrolled move to systemfsoftware 8d07118079, which is main after systemfsoftware#742 added ci-wall-clock. This repository belongs to the systemfsoftware organization, so the org-wide pack applies. The one changed file is .compound-engineering/config.yaml: ref: moves to the full SHA and pack: gains ci-wall-clock. The git source and path: are unchanged.

What the repin changes in the packs already enrolled

The new ref also brings the retained packs' current text:

• boundary-testing gains port-fakes-pass-the-port-contract.
• cell-architecture loses ports-separate-from-layers and changes handle-state-privacy and service-and-layer-boundaries.
• schema-laws changes rich-type-over-foreign-encoded.
• package-topology changes condition-branch-agreement and its README.

No file in this repository cites the removed rule.

The 1146 plan still records the old pin and "4 roots, 0 errors" (docs/plans/2026-10-10-1146-feat-effect-api-extractor-plan.md:317, :319, :446). This PR supersedes that record; it is not a reason to repin back to df908ddfe4. The plan is left unedited because this PR changes one file only.

Resolver

I ran packs-resolve.py (compound-engineering 3.30.3) in this checkout:

• --declared-only: exit 0, errors: [].
• Full resolve: exit 0, errors: [], warnings: []. It returned five roots, all at 8d071180798f9ea82ef5a4fd877152dacc02f6b6: cell-architecture, boundary-testing, schema-laws, package-topology, ci-wall-clock.

Review receipt

Run 20261012-205646-03f29a7f: ce-code-review mode:agent base:origin/main depth:full on head ba32e821f4 (base 9bd73962b7). Final disposition: Ready to merge; nothing blocks it at this head. The CE report step emitted "Ready with fixes" with an empty actionable list. The validator rejected finding 3. Finding 5 (P3) is a stale row in the frozen 1146 plan that this one-file PR cannot carry. It is escalated with #6's finding 3 as root decision D2 (see #6), and under REV1 only P0 and P1 findings block.

Packs loaded: all five above. Rule files read: cell-architecture 15, boundary-testing 6, schema-laws 10, package-topology 7, ci-wall-clock 12 (50 in all; READMEs are descriptions).

Only one rule matches this diff: ci-wall-clock/pr-gate-runs-only-what-the-diff-affects, through its review-invocation clause. It complies, because this review ran on the full path with every declared pack loaded. The other 49 rules do not match.

Constitution: systemfsoftware CONSTITUTION.md at 8d07118, which is identical to 02bf965, outranks the pack rules and AGENTS.md. No law is violated. CONST-T12 (applicability by meaning) decided the per-rule matching.

Reviewers: correctness, project standards (AGENTS.md plus the constitution), agent-native, learnings (the per-rule receipt above), and adversarial.

Cross-model: no peer CLI is installed, and Grok routes are excluded. The adversarial pass ran through omp's adversarial-reviewer agent. Its session log shows 18 turns served by MiniMax-M3.1-Flash-Preview, 21 by deepseek-v4.1-flash, and none by Claude or Grok. Every reviewer in this run was served by MiniMax or DeepSeek.

Findings and rulings:

#	Sev	File:line	Finding	Ruling
3	P3	.compound-engineering/config.yaml:5	AGENTS.md does not tell agents to review CI-facing diffs on the full path	Rejected by the validator; it would also need a second file in this PR
5	P3	docs/plans/…1146…-plan.md:446	The plan's R4 gate still reads 4 roots	Escalated as D2, non-blocking. The plan is frozen, so only a new dated plan declaring supersedes: can correct it, and that is a second file this PR does not carry. The root decides between a separate plan-supersession PR and these PR bodies as the record (CONST-S2).
1, 2, 4	P2/P3 advisory	config.yaml:3, :5	Repin changes retained rules; this repo's own CI already breaks one-stuck-job-backstop; packs enforce nothing off the full path	Residual risks, stated above or carried by the next CI-facing PR's review

Privacy

Before push, I ran a whole-word grep for every private systemfsoftware repository name (54) over the diff, the commit message and this body. The only match was AGENTS.md, which is also a common file name, and every hit is this repository's own AGENTS.md file. A grep for host paths matched nothing.

Re-verification (2026-10-13, head ba32e821f4)

• Scope: one file, .compound-engineering/config.yaml; head ba32e821f4152e4a8ed93957756afada06ef8daa, unchanged since the receipt above.
• Resolver (packs-resolve.py, compound-engineering 3.30.3): --declared-only exit 0 with no errors. Full resolve exit 0 with no errors or warnings, five roots at 8d071180798f9ea82ef5a4fd877152dacc02f6b6.
• Privacy: re-ran the whole-word grep against a fresh list of all 54 private systemfsoftware repository names, fetched today. The diff and commit message matched none. This body matched only AGENTS.md, which is this repository's own file. No host paths matched.
• CI: CI run 38312651873 passed all 11 jobs, and Changeset Check run 38312652042 passed.

ce-compound

ce-compound was evaluated for this PR (mode non-interactive) and ended Documentation skipped. Reason: the enrollment is a config repin with no solved problem behind it. Its one lesson, that a repin can silently change rules in the packs already enrolled, is recorded above under "What the repin changes". Nothing yet detects such a change, so there is no fix to document.

[ryanleecode]

Root verdict: one file, full 40-char pin 8d071180798f, resolver five roots with no errors or warnings, 12/12 green on ba32e82, 0 behind, 0 threads, MiniMax/DeepSeek peer with no Grok, constitution ruled. Compound skip accepted for a one-file repin.