Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows
fix: use gh run watch instead of manual polling in pr-review Step 5 (#1246)
worktrunk-botc24cbdf22791
diff --git a/.claude/skills/pr-review/SKILL.md b/.claude/skills/pr-review/SKILL.mdindex 947176e65..5e39ddde3 100644--- a/.claude/skills/pr-review/SKILL.md+++ b/.claude/skills/pr-review/SKILL.md@@ -293,11 +293,24 @@ description: new text here### 5. Monitor CI-After approving, check whether CI has finished. Exclude the current workflow's-own check to avoid a circular wait (Claude polling itself):+**Skip this step** if the verdict was "stay silent" (self-authored PR with no+concerns). There is no approval to dismiss on failure, so monitoring adds no+value.++After approving, wait for CI to finish using `gh run watch` (consistent with+the `running-in-ci` skill). Exclude the current workflow's own check to avoid+a circular wait:++```bash+# Find the CI run triggered by this PR's HEAD commit.+RUN_ID=$(gh run list --branch <branch> --commit "$HEAD_SHA" \+ --workflow ci.yaml --json databaseId --jq '.[0].databaseId')+gh run watch "$RUN_ID" --exit-status 2>&1 || true+```++After `gh run watch` completes, verify final status:```bash-# $GITHUB_WORKFLOW is set in CI; when unset, no checks are excluded.gh pr view <number> --json statusCheckRollup \--jq '[.statusCheckRollup[]| select(env.GITHUB_WORKFLOW == null@@ -307,11 +320,6 @@ gh pr view <number> --json statusCheckRollup \```- **All checks passed** → done, no further action.-- **Checks still running** → poll until complete (sleep 30–60s between checks).- Polling is intentionally unbounded — CI compute is cheap and catching failures- before the author looks at the PR is more valuable than saving a few minutes- of polling. The results are always visible on the PR regardless, so the worst- case of a long poll is wasted compute, not missed information.- **A check failed** → if it's a flaky test or unrelated infrastructurefailure, no action needed. If the failure is related to the PR changes:1. Investigate the failure and post a follow-up review (COMMENT) with