babysit-pr

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

5|1|Updated Jun 26, 2026
One-click install
npx skills add https://github.com/looooonk/better-codex --skill babysit-pr-looooonk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: babysit-pr
Source: https://github.com/looooonk/better-codex/tree/main/.codex/skills/babysit-pr
Command: npx skills add https://github.com/looooonk/better-codex --skill babysit-pr-looooonk

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? After opening a pull request, developers must repeatedly check CI results, watch for new review comments, diagnose flaky failures, and track mergeability. This Skill automates that entire monitoring loop so the PR is watched continuously until it is merged, closed, or genuinely blocked. ## Core Features & Use Cases - Continuous PR Watching: Polls CI checks, workflow runs, review comments, and merge state via a watcher script with one-shot (--once) or streaming (--watch) modes. - CI Failure Diagnosis and Retry: Classifies failures as branch-related or flaky using failed-job logs, auto-fixes branch-related issues, and reruns flaky jobs up to 3 times per SHA. - Review Feedback Handling: Surfaces published review comments from trusted humans and approved bots, applies actionable fixes, and resolves threads only under a strict GitHub mutation policy. - Use Case: After pushing a branch and opening a PR, ask the agent to babysit it; it will watch CI, fix a failing lint job, push the fix, address a reviewer's comment, and keep polling until the PR merges. ## Quick Start Ask the agent to babysit the pull request for the current branch and keep watching its CI and review comments until it is merged.

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 automatically?▼

Run the watcher script with --watch to stream JSONL snapshots of CI checks, review comments, and mergeability state. It accepts a PR number, URL, or auto-detection from the current branch, and keeps polling until the PR is merged or closed.

How to retry flaky GitHub Actions failures on a PR?▼

Use the watcher's --retry-failed-now mode, which reruns only failed jobs via gh run rerun --failed. Retries are capped at 3 per head SHA, and the watcher only recommends reruns when checks are terminal and failures look flaky or unrelated to the branch.

Does the PR watcher handle pending GitHub reviews?▼

No, it intentionally ignores review submissions in GitHub's PENDING state and their inline comments. Pending feedback is not marked as seen, so it becomes eligible to surface only after the reviewer submits the review.

Can the agent reply to human review comments automatically?▼

No, it never posts replies to human-authored review threads without explicit user confirmation of the exact response. It can fix actionable feedback in code, push commits, and resolve threads only from the requesting user or approved bots.

What are the limitations of automated PR babysitting?▼

It stops and asks for help on infrastructure outages, exhausted flaky retry budgets, permission failures, dirty worktrees, or ambiguous reviewer requests. It also avoids patching unrelated flaky tests, CI configuration, or dependency issues not caused by the branch.