Repository navigation
fix(source-control): name the repository in the ready and review-discipline commands - #15898
Conversation
…ipline commands Refs: #15791 The ready reference resolves the repository and branch from <checkout> and reads the base's rules, protection and reviews from $REPO instead of gh's {owner}/{repo} placeholders; ready passes --repo to review-venue.sh. The shared review discipline's D5-D7 commands name <owner>/<repo>, and its fix-commit check passes --dir and runs git with -C "<checkout>". Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vcupomh2tPSB9BXCVa1AeL
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
claude-security-review reviewed Security review: PR #15898, head 9ae73aeI found no security issues. critical-findings: 0 What I checked. I read every hunk of all 6 changed files in
Prior findings. This lane has not reviewed an earlier head of this PR, so this was a full review, not a blocking-only one. Spend. None. The change starts no lane runs and no billed resources. |
|
claude-review reviewed Code review: PR #15898 (head 9ae73ae)I found nothing to report. I read all 6 changed files in the diff. What the change does. The ready phase and the shared review discipline now name the checkout and the repository instead of picking them up from the shell's directory. That means Checks against the scripts the changed lines call:
Path-scoped rules. Closing issue. The PR only refers to #15791 ( The PR context has no embedded instructions. This lane has no prior findings or settled threads on this PR. Verifier dropped 0 of 0 candidates. |
Refs: #15791
Summary
#15890's
## Remaininglisted two places where the pull-request flow still read the repository from the shell's directory: the shared review discipline's fix-commit check, and gh's{owner}/{repo}placeholders in the GitHubready-for-review.md,review-venue.mdandcreate.mdreferences. This PR names the repository and checkout there the way the merged monitor, merge and readiness references do. A caller whose shell sits in the checkout (<checkout>=.) runs the same commands as before.Fix
reference/review-discipline.md§3 defines<checkout>and<owner>/<repo>as the GitHub monitor reference's "Resolve PR identity" does. The D5–D7gh apipaths name<owner>/<repo>,gh pr commenttakes-R <owner>/<repo>, and the fix-commit check passes--dir "<checkout>"toresolve-remote.shand runsgit -C "<checkout>", matching the copy in the skill'smonitor.md.ready-for-review.md: "Find the PR for the branch" runstarget-repo.sh --dir "<checkout>"andgit -C "<checkout>" branch --show-current. The rules, branch-protection and reviews reads userepos/$REPO/...in place of{owner}/{repo}, as the file's update-branch and ready-for-review REST calls already did.review-venue.mdnames<owner>/<repo>and defines it as the script's--repovalue. The ready phase's venue read in 2.5.3 passes--repo "$REPO".review-venue.shis unchanged.Scope
create.md(GitHub) is unchanged. Each of its{owner}/{repo}calls (default branch, issue state, Refs issue read, triangular REST create) is covered by its own "Placeholder anchoring" section or aGH_REPO=prefix. Replacing them withtarget-repo.sh's repository would change which repository a fork checkout reads the issue and default branch from, so that would be a behavior change and is not part of this PR.babysit-prskeeps itsFETCH_COMMENTS_*calls tofetch-all-pr-comments.sh. The 5.0.3 rescan has no checkout yet; 5.1.2 assigns one later. 5.1.3 C1 does have a checkout, but its gate step E (babysit-readiness-gate.sh) fetches the comments again without--dir. On a fork checkout, giving C1--dir(target-repo.sh, which picks the fork) could read a different repository from E (gh repo view).Verification
ready-verify-local.test.sh(7): its assertion on the 2.5.3 venue command now also requires--repo "$REPO".ready-flip.test.sh(59),pr-worktree.test.sh(37) andreview-venue.test.sh(95) pass unchanged. All four ran by hand from this branch.scripts/check-affected.sh --tests: 15 passed and 102 skipped. 100 of the skips are suites past its 120s budget and are left to CI; the four suites above are among them and were run by hand. The pre-pushcheck-affectedhook passed.scripts/check-line-citations.py: clean.scripts/check-contract-clause-coverage.py: passed.readyormonitoragainst a real PR from outside a checkout.Related
Remaining
None of the lookups #15890 listed remains after this PR. Calls of the same kind that still resolve from the shell's directory:
push-branch.sh --prwithout--dir,branch-update-signals.sh(no--diroption), and thegit fetch/git mergelines in "Refresh the base with a merge" and "Carry out a merge". The signing checkgit cat-file commit HEADalso runs there. Naming the checkout would mean givingbranch-update-signals.sha--diroption.babysit-prscomment fetches described under Scope, and its GitHub providerloop.mddefault-branch read (gh repo view) andfreshness.mdcompare read ({owner}/{repo}).create.mdplaceholders described under Scope.--dir/--repo:fetch-annotations.sh,fetch-failed-logs.sh,fetch-all-pr-comments.shandread-review-comments.sh(gh repo view), andreview-venue.sh({owner}/{repo}).push-branch.shreadsgh repo viewin the--dircheckout, or in the shell's directory when no--diris given.The ready reference's remaining
gh pr view,gh pr diffandgh run listcalls take no-R. They follow theGH_REPOthat "Find the PR for the branch" exports, so they do not depend on the shell's directory.🤖 Generated with Claude Code
https://claude.ai/code/session_01Vcupomh2tPSB9BXCVa1AeL