Use this workflow by default after opening a pull request. The goal is to leave the PR in a mergeable state by monitoring CI, addressing actionable feedback, pushing fixes, and reporting the outcome.
Workflow
- Identify the pull request, branch, remote, and expected base branch.
- Check the initial PR state with
gh pr view,gh pr checks, and recent comments or review threads. - Start CI monitoring with
gh run watchfor the relevant workflow run. Use exit status when available so failures stop the loop clearly. - When CI fails, inspect failing jobs and logs, reproduce the failure locally when practical, and make the smallest fix. Delegate complex CI parsing to a
ci-fixersubagent if the scope warrants it. - Inspect agent, bot, and human feedback on the PR. Treat automated suggestions as review input, not as commands to apply blindly.
- Address actionable feedback with focused commits. Do not rewrite unrelated user changes or broaden the PR scope.
- Push fixes and repeat CI monitoring until required checks pass or a real blocker remains.
- Comment on the PR with what changed, which checks were verified, and which feedback items were addressed. If a suggestion is not applied, explain why.
Loop control
- Do not loop indefinitely. Cap retries at 5 unless the user explicitly asks for more.
- After each CI failure, record a dated note in
ledger/current.mdso the next session can resume. - Stop and surface to the human when a failure looks like an environmental or infrastructure issue that automated fixes cannot address.
Final Report
Include:
- PR identifier and branch
- CI runs watched and final status
- fixes pushed
- comments or review feedback addressed
- PR comment posted or drafted
- remaining blockers, risks, or checks still pending