open-and-monitor Skill | Agent Skills

open-and-monitor

Arguments: $ARGUMENTS (passed through to draft-pr).

Execute this flow fully, stop only if blocked.

1) Open draft PR

Invoke draft-pr with $ARGUMENTS.

Capture the PR number and URL:

gh pr view --json number,url --jq '.number, .url'

2) Monitor checks and bot comments (fail-fast loop)

Poll checks and bot comments together. Max 40 iterations, 30s apart (~20 min). On each iteration:

Query check status:

gh pr view <PR> --json statusCheckRollup --jq '
  .statusCheckRollup[] | [.name, .status, .conclusion] | @tsv
'

Query bot comments:

gh pr view <PR> --json comments,reviews --jq '
  [.comments[], .reviews[]]
  | map(select(
      .author.is_bot == true
      or (.author.login | test("bot|copilot|dependabot|renovate|coderabbit"; "i"))
    ))
  | length
'

Evaluate:
- Any check FAILED? Get the changed files list (gh pr diff --name-only) and determine if the failure is relevant to PR changes (touched files, tests, lint, typecheck, build, or logic connected to changed code). If relevant, break out of the loop immediately and go to step 3.
- Bot comments appeared? Break out immediately and go to step 3.
- All checks PASSED and no bot comments? Skip to step 5.
- Checks still PENDING, nothing failed yet? sleep 30 and continue polling.
If 40 iterations pass with checks still pending, proceed to step 3 with whatever state exists.

3) Fix

Invoke fix-pr. It handles failing checks, bot/reviewer comments, commits, pushes, and replies.

4) Verify fixes

After fix-pr completes, re-enter the monitoring loop from step 2 (same fail-fast logic). If a relevant failure or new bot comment appears again, invoke fix-pr one more time (max 2 total invocations). After the second attempt, proceed to step 5 regardless of outcome.

5) Final report

Output a summary:

## PR Summary

- **PR**: <URL>
- **Checks**: passed | failed
- **Bot comments**: <count>
- **fix-pr invocations**: 0 | 1 | 2
- **Status**: clean | needs attention

If checks are still failing after 2 fix-pr attempts, list the remaining failures and recommend next steps.

Agent Skills: open-and-monitor