Consolidate capability qualification and camera/workbench experiments

[Tom-R-Main]

Footwork's capability qualification and independently developed camera, desk inbox and physical workbench now share one reviewed integration branch. This brings both requested work streams onto main while retaining their experiment records and qualification limits.

The integration repairs two fresh-checkout problems: physical controller sockets now use a short, private directory independent of checkout length, and frozen onboarding packets verify their wheel against the requested immutable source commit. CI now exercises the camera, inbox, workbench and pinned MuJoCo simulation suites in addition to the existing Python, Node, Rust and wheel checks.

The original macOS capability candidate remains byte-identical: ZIP 8dc81c724bf617a9f3aa4833f250d5124b9e8c5e9a1c504c67d634080f44fd1f, qualified product 445a348ff31243c2f3d90baae9ce209d89bd7eac, release source d9315593886c17196db477db969f21481fc10e8d. A newly built wheel from this branch includes experimental artifact checking and is a separate artifact. The prerelease will use the exact qualified candidate.

Validation:

• Python source: 1,594 passed, 27 skipped, 12 browser tests deselected after remediation; 105 final changed-file/notebook checks passed.
• Node: 97 passed; synthetic camera: 18 passed; inbox/workbench: 30 passed plus 11 subtests.
• Packet/release verification: 44 tests passed; both original packet and release ZIP verify unchanged after consolidation. Single-snapshot parsing/hashing, unsafe ZIP metadata and non-commit source objects have regressions.
• Controller: 18 process scenarios and 10 socket regressions passed; AgentBench: 15 passed. Full simulation reruns passed 314, then 330 after remediation; the final two producer-identity regressions and 42 affected simulation checks also pass. Current hosted Linux CI will exercise the complete final suite before merge.
• Ruff, notebook validation and whitespace checks passed. One fresh Opus 5.5 static review found no product-path blocker. Its actionable findings were repaired and independently scrutinized; the review and resolutions are recorded in jevdual/results/consolidation-20261014/. The remaining fail-closed limits are explicit.

The initial failed fresh-checkout runs and Linux CI run are retained. Linux CI initially failed only because synthetic AX tests imported the macOS-only framework; an explicit Linux fake now preserves those tests and refuses unconfigured actions. A bounded publication audit found only metadata for the live camera/desk lane and controlled simulation screenshots; private note/camera inputs remain excluded. Private run evidence and the unrelated untracked draft are preserved separately before checkout retirement.

Qualification remains narrow: guest permission reset, installed dependency checks and actual agent execution are separate evidence lanes. Simulation does not qualify physical hardware. This PR does not claim another Mac, a never-granted OS, arbitrary target discovery, Screen Recording or saved-file persistence. README-only Codex adoption on the existing Mac follows landing, with the VM stopped and the requested memory headroom.

Merge with a true merge commit so the qualified product and release-source commits remain reachable from main. The final remediation head is 881f27d; merge requires its current CI and inspection of every review thread.

A second Linux-only failure came from a test clock modifying the shared standard-library time module. The test now uses an isolated clock; its new regression failed before the change and all 57 native fixture tests pass afterward. Runtime/helper bytes and the qualified package are unchanged.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-14T17:23:36.079789Z	57ed816	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.