[rabii-chaarani]

Summary

Fixes the source-layout cause of failed Release job 122023180345.

• Delete all four loose Python evidence helpers. Preserve their exact bytes and original paths in a deterministic historical archive, with hashes and reconstruction notes.
• Run the unchanged repository release gate before packaging in CI. Add one workflow-policy regression so the step cannot silently disappear.
• Keep Rust-only rejection, production confirmations, historical cleanup reports and exact-SHA release identity unchanged.

Authority and tracking

• COD-99 · CodebaseGraph workspace / Maintenance.
• User approval on 2026-10-27: “i approve archiving the python helpers”. This authorizes archival, not release replacement or publication.
• Existing contracts: release process at failed source; canonical wiki knowledge:architecture/graph-disk-cleanup-contract (DC-S1–DC-S5 rev 0.1 / approved amendments rev 0.2). No duplicate spec, acceptance change or new architecture claim.
• Scryer ledger chg-cxz5az; existing claims resp-14 / resp-15 remain unchanged.

Executed acceptance evidence

Reviewed/tested source: fa0c6bccde1df65c747dbf853a455abb29be0191; local Darwin/aarch64, Rust 1.96.0.

• Archive generated twice with identical bytes; safe exact member set; all four extracted helper hashes equal original Git inputs.
• cargo fmt --check — passed.
• cargo test -p xtask --locked — 43 passed, 0 failed/ignored.
• cargo nextest run -p xtask --locked --config-file /private/tmp/codebasegraph-cod99-verification/nextest.toml --profile cod99 — 43 passed, 0 skipped; JUnit retained.
• cargo clippy -p xtask --all-targets --locked -- -D warnings — passed.
• Workflow checks, ordinary release gate and confirmed-production gate — passed; production gate passed again after commit.
• Failure probes still reject a loose .py in the original evidence directory and production without both external confirmations.
• Fresh ordinary read-only reviewer: openai-codex/gpt-6.1-sol, thinking="max"; no confirmed defects. Reviewer independently ran the affected checks and verified archive bytes/index inclusion.

Durable evidence and review · Archive manifest.

Remaining gates and limits

Full product/native-platform matrix and package dry-run are pending PR CI, not inferred from local xtask tests. Historical Python helper tests were not rerun. No constants, runtime or package payloads changed.

The shared Scryer model targets the unrelated older original worktree, not this isolated fix; no JUnit verdict was activated against mismatched source and no broad drift reset occurred. Evidence is retained for correct-source synchronization.

Rerunning the original job will still check immutable 7038e66 and fail. Publishing corrected source requires a separate trusted replacement release PR/version decision and its own successful exact-SHA CI. This PR does not move tags, modify #156 labels, dispatch publication or close maintainer-owned aggregate/release gates.

[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 2026-10-27T05:53:59.301912Z	fa0c6bc	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.