Does it work with any GitHub repository?

It works with any repository where the user has the GitHub CLI (gh) configured and appropriate permissions to push branches and trigger workflow runs.

Does it support mobile development environments?

Yes, this skill is specifically optimized for Termux on Android as part of the codex-termux-pocket suite, but it functions in any standard terminal environment.

Can it fix code based on reviewer comments?

Yes, it identifies actionable review comments from authorized users and bots, applies local patches, and pushes the changes directly to the PR branch.

How does PR Babysitter handle flaky tests?

The skill classifies failures by inspecting logs via the GitHub CLI and provides up to 3 automated retries for failures identified as transient or infrastructure-related.

When does the monitoring loop stop?

Monitoring stops when the PR is merged or closed, when it reaches a 'ready to merge' state (all green and approved), or if a blocker requires manual human intervention.

PR Babysitter

Name: PR Babysitter
Author: Kbediako

byKbediako

•

Collaboration Tools

Automates the monitoring and management of GitHub pull requests by diagnosing CI failures, retrying flaky tests, and addressing review comments until merge.

PR Babysitter is a high-autonomy workflow skill that eliminates the manual 'wait-and-watch' period after creating a GitHub pull request. It continuously polls the PR status, intelligently classifies CI failures into code-related or flaky categories, and takes proactive action by either patching code or triggering retries. By monitoring review comments and mergeability in real-time, it ensures that your PR stays on track for integration without requiring constant manual check-ins, stopping only when the task is successfully completed or requires specific human decision-making.

Key Features

01Automated retry logic for transient test failures with a configurable budget.

02Proactive code patching and pushing in response to actionable reviewer feedback.

03Intelligent failure classification to distinguish between code bugs and flaky infrastructure.

044 GitHub stars

05Live merge-conflict monitoring and mergeability state tracking.

06Continuous real-time polling of CI/CD workflow runs and GitHub Actions status.

Use Cases

01Automating the 'fix-and-push' cycle for minor code review feedback and linting errors.

02Handling flaky CI environments that require multiple reruns to achieve a green state.

03Managing long-running CI suites where manual monitoring is inefficient.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add kbediako/codex-termux-pocket babysit-pr

For use in Claude.ai and ChatGPT

Download Skill