babysit-pr

Monitor GitHub pull request CI, review comments, and mergeability until merge or closure.

5|1|Updated May 5, 2026
One-click install
npx skills add https://github.com/cbusillo/codex-skills --skill babysit-pr-cbusillo
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: babysit-pr
Source: https://github.com/cbusillo/codex-skills/tree/main/skills/babysit-pr
Command: npx skills add https://github.com/cbusillo/codex-skills --skill babysit-pr-cbusillo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? After pushing a PR, engineers repeatedly check CI status, review feedback, and merge conflicts by hand. This Skill automates that follow-through by continuously polling a GitHub pull request, diagnosing failures, retrying flaky checks, and surfacing new review comments until the PR is merged, closed, or needs human help. ## Core Features & Use Cases - Continuous PR Watching: Streams JSONL snapshots of CI checks, review state, and mergeability via a watcher script, with adaptive polling (1 minute active, 5 minutes for quiet green PRs). - CI Failure Diagnosis and Retry: Classifies failures as branch-related or flaky using failed-job logs, auto-fixes branch-related issues, and reruns flaky failures up to 3 times per head SHA. - Review Comment Handling: Surfaces issue comments, inline review comments, and review submissions from humans and bots, fixes actionable feedback, and resolves threads after pushing. - Use Case: After pushing review fixes to an open PR, ask the agent to keep watching; it restarts the watcher, confirms the new CI round, surfaces any late review comments, and reports when the PR is ready for a merge decision. ## Quick Start Use the babysit-pr skill to monitor the current branch's pull request, including CI checks, review comments, and mergeability, until it is merged or closed.

Frequently Asked Questions about babysit-pr

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I monitor a GitHub pull request until it merges?

Run the watcher in continuous mode with uv run scripts/gh_pr_watch.py --pr auto --watch. It streams JSONL snapshots of CI, review, and mergeability state and keeps polling until the PR is merged, closed, or blocked on user help.

How do I automatically retry flaky CI failures on a PR?

The watcher emits a retry_failed_checks action when failed runs look flaky and rerunnable. Trigger reruns with uv run scripts/gh_pr_watch.py --pr auto --retry-failed-now, limited to 3 retry cycles per head SHA before stopping for user help.

Can this skill merge a pull request automatically?

No. The skill never merges a PR; it only reports when a PR is ready for a merge decision. Merge execution is delegated to the separate github skill and requires explicit merge authorization plus a fresh readiness check.

Does the PR watcher handle review comments from bots and humans?

Yes. It surfaces issue comments, inline review comments, and review submissions from both humans and automation bots. Actionable feedback can be fixed and pushed, but replies to human-authored comments require explicit user confirmation before posting.

Why does the watcher report check_evidence_incomplete?

It means REST check counts cannot prove a terminal CI round, due to lower-bound counts, unavailable components, or a head SHA mismatch. The watcher keeps polling and will not rerun failed jobs from that incomplete evidence alone.

What are the limitations of automated PR babysitting?

It stops for user help on infrastructure outages, exhausted flaky retry budgets, permission failures, ambiguous reviewer requests, or protected PR head branches. It also never force-pushes, never patches unrelated flaky tests, and does not merge PRs itself.