What does this skill do, and when should you use it?
This is a skill focused on monitoring a pull request (PR) throughout its lifecycle. It continuously checks the PR status, CI checks, comments, and review threads, fixing real issues until the PR is ready to merge. The skill uses GitHub CLI and GraphQL API to fetch detailed PR information and unresolved threads, and follows strict rules to ensure everything is clean before reporting. The skill comes from the claude-mem repository, which contains multiple skills, but this one operates independently.
The skill: 1) uses gh pr view to get coarse PR status including branch, mergeability, checks summary, etc.; 2) uses GraphQL queries to fetch unresolved review threads with pagination for large numbers; 3) polls at a practical interval (default 30-60 seconds) while checks are pending; 4) reads new comments and threads, distinguishing bot summaries from actual code issues; 5) creates focused commits for real problems, runs tests, and pushes; 6) resolves stale threads only after verifying the fix; 7) stops when all checks pass (or are intentionally skipped), review decision is acceptable, no actionable comments remain, and no unresolved threads exist.
- A developer needs to keep an eye on an important PR and wants to avoid manually refreshing while CI runs
- A maintainer wants to ensure all review comments are addressed before merging a PR
- A team wants to automate handling of common PR issues like stale threads or bot reports
- Before a release, you need final verification that all checks pass and no threads are unresolved
How do you install this skill?
- Executing this skill requires gh CLI and GitHub network access, which may be inaccessible directly in mainland China; consider using a proxy or alternative tools.
- The skill automatically resolves review threads, which may erroneously resolve threads that are not fully fixed; recommend adding manual confirmation before critical operations.
- Static assessment did not execute actual commands; command correctness should be verified in a real environment.
- The skill does not explicitly handle GitHub API rate limits or authentication failures, potentially leading to unclear error feedback.
- Shell / CLI
- Network access
GitHub CLI (gh)jqGit
This skill is part of the claude-mem repository but can be used independently. Clone the repo and copy the plugin/skills/babysit/ directory to your skills folder, or extract its content as a standalone skill. Requires GitHub CLI (gh) and jq.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/thedotmack/claude-mem.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/plugin/skills/babysit" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Please babysit PR #123 until all checks pass and review comments are resolved.
Directly describe the task to a skill-capable AI client, e.g., "Please babysit PR #123 until all checks pass and review comments are resolved." The skill will identify the PR number, poll status, handle comments, and commit fixes. Ensure you are logged into GitHub CLI (gh auth login) in your terminal and have appropriate permissions on the repository.
What are this skill's strengths and limitations?
- Detailed copy-paste commands using GitHub CLI and GraphQL
- Clear stopping criteria to avoid premature reporting of completion
- Pagination logic for many review threads, suitable for large PRs
- Emphasizes verifying fixes before resolving threads, reducing mistakes
- GitHub-only; doesn't support GitLab or other platforms
- Requires installation and configuration of GitHub CLI and jq
- No mention of automated tests or cross-platform verification in the docs
- Fixed polling interval may not suit urgent scenarios requiring immediate response
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| PR Babysitter: Watch Pull Requests Until Merge this page | 47 · Use with care | ★ 98k | 3d ago | Apache-2.0 |
| PR Babysitter ✓ OpenAI · Official | 57 · Use with care | ★ 128k | 3d ago | Apache-2.0 |
| babysit — PR Babysitting Skill | 54 · Use with care | ★ 8.7k | 1d ago | Apache-2.0 |
| Ship a PR (e2e repo PR delivery workflow) | 59 · Recommended | ★ 8.7k | 1d ago | Apache-2.0 |
| CodeRabbit Autofix | 53 · Use with care | ★ 188 | 4d ago | MIT |
How did FollowSkills review this skill?
The skill monitors and modifies PR review threads via GitHub CLI and GraphQL API, performing read operations and one write operation (resolving a review thread). Permissions are minimal: it only calls gh api, relying on user's prior authentication; the write is explicit and targeted. Data flow is transparent: commands are explicit and output displayed locally. Static review cannot verify safe execution, but code does not show dangerous patterns (e.g., shell injection) and follows security best practices (array-based arguments). Deduction: lacks explicit confirmation steps; the skill auto-resolves review threads, requiring user trust in the skill's judgment. Publisher not verified, but the skill itself does not exhibit undue permissions.
The skill's instructions are internally consistent, providing clear steps and shell scripts for queries. But it depends on external tools (gh) and environment (GitHub repo), has no built-in error handling or feedback mechanism; scripts assume gh is installed and authenticated, no handling of auth failures or rate limits. As static review, actual execution cannot be verified. Deduction: no error handling, no debug output, failure feedback not clear.
The skill's scenario is clear: monitor a PR until it can be merged. Trigger conditions are clear (when user asks to babysit). However, the skill relies on GitHub CLI and GraphQL API, which may be inaccessible directly from mainland China, affecting feasibility. Boundaries are not clear (e.g., number of PRs, processing limits). Deduction: no consideration for mainland China network reachability, and no clear definition of boundaries beyond monitoring termination conditions.
The repository has clear Readme, license (Apache-2.0), versioning (package.json version), CI, security policy, and contribution guidelines. The skill file itself is well-structured with description and workflow. Deduction: the skill file does not include version or changelog, no separate troubleshooting section, but repository-level docs are sufficient.
The skill is designed to accomplish the core task of monitoring PRs, but actual effectiveness cannot be verified statically. It provides concrete commands and output processing; expected output is a list of unresolved threads, but it is not verified if directly usable. A score of 4 is given based on implementability, but deducted for lack of execution verification and output format not clearly indicated.
The skill file itself has no tests. The repo has CI and test suites, but not for this specific skill. Static review cannot verify command correctness. Deduction: no evidence of third-party verification, and key behaviors cannot be reproduced.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →