[echoomegaprime]

Purpose

Focused four-file Finisher observer/collector implementation. Captures an unsigned, fail-closed journey over eight agent fronts with exact owner readbacks, safe process cleanup, rollback restoration checks, documentation, and regression tests.

Dependency / base

• Stacked on draft Repair release checks, readiness deadlines, and dependency policy #25 — branch agent/certforge-readiness-20261003, pinned base SHA 424fc76fa09d9a83fcbe0d2ec12dc62ef97859c9.
• Repair release checks, readiness deadlines, and dependency policy #25 provides required production-E2E schema/entry points. Do not merge independently or before Repair release checks, readiness deadlines, and dependency policy #25 is integrated and revalidated.
• Replaces the contaminated original draft Finisher: independent fail-closed live verification collector (task 50135) #44, which remains unchanged; PR47 excludes all 16 unrelated inherited commits.

Exact source and scope

• Current head: 2e0d592b031c5482bf69f150aec74012dc953bf8.
• Git tree: 5efe62b70c08a911e2bf7ae9df3d766506215ca7 — bit-for-bit match to isolated QUENCH hardening commit f5a8fbabb2ee4c70b0934a99631b631e0809c74d.
• Four files relative to Repair release checks, readiness deadlines, and dependency policy #25:
  • docs/FINISHER_COLLECTOR.md
  • scripts/collect_finisher_e2e.py
  • src/echo_certification_forge/finisher_collector.py
  • tests/test_finisher_collector.py
• Fourth commit strengthens trust boundaries: 24 separate named owner-issued/recalled stability-front observations (three windows × eight fronts), exact route/provider/snapshot hash binding while permitting one shared Cloudflare singleton run ID, and exact comparison to four owner-approved immutable rollback checkpoint/config digest pins. Malformed/replayed/cross-front proofs fail closed.
• No generated P1 report, deployment asset, or unrelated file is part of the diff.

Checks (2026-11-08)

• Current hardening tree: QUENCH Windows Python 3.14 focused tests 74 passed, 6 skipped, exit 0; Python compile and whitespace check passed.
• Current hardening tree: P1 acceptance 3/3 scenarios passed, exit 0. Source-only and tampered evidence remain blocked; deploy gate denied and verdict NOT_READY.
• Previous pre-hardening tree only: QUENCH full pytest -q 760 passed, 9 skipped; FORGE Linux focused 65 passed. These results are historical and do not qualify the new head.
• Current head GitHub-hosted full CI, P1/P6/P7 acceptance and CodeQL: pending independent results. Local full suite was intentionally deferred due low QUENCH available RAM and bounded FORGE I/O stall, not represented as passing.

Release disposition: NOT_READY

No admitted Finisher-specific independent eight-front production attestation, source/artifact identity sign-off, rollback operational proof, or authorized independent signed release verdict exists for this PR. Keep DRAFT and prevent deployment/merge until hosted CI, CodeQL, independent security review, real owner admission, production E2E, and sign-off are complete. Do not substitute a fixture or simulated success for genuine evidence.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

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

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-12T02:22:11.455632Z	2e0d592	Draft marked ready

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.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2e0d592b03

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Restore the singleton when front-proof validation fails

If an owner probe completes the front journey, including its terminal halt, but the returned proof fails _front validation (for example, an earlier status snapshot has aged beyond the 30-second freshness limit), the exception skips _restoration entirely. The CLI then reports HELD and exits, leaving the shared singleton halted even though restoration was admitted. Make recovery independent of successful proof validation, while retaining fresh admission checks and the failed collection outcome, so an evidence-validation failure does not leave the target unavailable.

Useful? React with 👍 / 👎.