Skip to content

fix(ce-sweep): scope merge evidence to full repository identity - #1994

Merged
tmchow merged 1 commit into
mainfrom
t3/evaluate-community-pr-1992
Nov 4, 2026
Merged

tmchow merged 1 commit into
mainfrom
t3/evaluate-community-pr-1992

Conversation

@tmchow

@tmchow tmchow commented Nov 4, 2026

Copy link
Copy Markdown
Collaborator

ce-sweep now requires a PR candidate to match the checkout’s full host/owner/repository identity before using it as merge evidence. This addresses the P1 found after #1992, where a GitHub.com issue could be falsely resolved using an unrelated, same-numbered Enterprise PR.

Follow-up to John Costa’s #1992. Source reads and checkout verification use explicit host-qualified repositories; unresolved identity preserves the item. The fixture now distinguishes same-host successes from foreign-host closing PRs and stored-number collisions on both open and completed issues.

Validation

  • bun run test: passed after the runner’s automatic serial retry recovered eight subprocess timeouts across five unrelated ce-work files.
  • Skill guards, focused sweep/catalog tests, typecheck, release validation, plugin validation, and git diff --check: passed. The new explicit-repository command guard fails against merged main.
  • Live gh checks confirmed host-qualified repository resolution and PR reads. No live Enterprise mutation was performed.
  • Before-change Codex reproduced the incorrect merge SHA and false closure. Claude and Grok compensated for the old prose in that run.
  • Final fresh-agent scenario: Claude and Codex passed. Grok failed, using a foreign stored PR for issue 113 and retaining an old fix claim on issue 102. Earlier Grok attempts also varied on plan reconciliation. Moving the identity gate beside the command did not eliminate that inconsistency. This is a remaining validation limitation, not a claimed three-host pass; do not treat this PR as merge-ready.

Security Disclosure

Prevents merge evidence from crossing GitHub host/repository boundaries in the skill’s verification contract. Adds read-only source repository resolution and explicit repository selectors; source-side close-out remains subject to standing approval and read-back. The Grok behavior failure above remains a risk for instruction execution.

Agent Disclosure

  • Model: T3 Code / Codex harness · gpt-6-astra

Co-authored-by: John Costa costajohnt@gmail.com

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 4, 2026 •

Copy link
Copy Markdown

Codex Review Summary

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

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-04T04:25:07.827012Z aee810f 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.

@tmchow
tmchow merged commit a3bcedf into main Nov 4, 2026
10 checks passed
@github-actions github-actions Bot mentioned this pull request Nov 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant