babysit

Monitors CI checks on a pushed branch, fixes failures, and pushes until the PR merges.

3|Updated Nov 8, 2014
One-click install
npx skills add https://github.com/mintuz/.dotfiles --skill babysit-mintuz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: babysit
Source: https://github.com/mintuz/.dotfiles/tree/main/agents/.agents/skills/babysit
Command: npx skills add https://github.com/mintuz/.dotfiles --skill babysit-mintuz

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Watching CI after pushing a branch is tedious: failures need diagnosis, fixes need pushing, and the cycle repeats until green. This Skill runs that loop autonomously so you can walk away and return to either a merged PR or one precise blocker. ## Core Features & Use Cases - Autonomous CI watch loop: Polls structured check status via gh (GitHub) or tea (Forgejo/Gitea) every 20-30 seconds, re-arming against the new head SHA after every push. - Failure classification: Classifies every red check as a branch failure, infra failure, or base failure, with bounded retries (3 strikes for infra, 3 fix attempts per check) before stopping and reporting. - Local-first diagnosis: Reproduces failures locally with the exact workflow command and runs the branch's changed test files during CI waits to pre-empt the next round. - Use Case: You push a feature branch, open a PR, and hand it to the loop. It applies an auto-merge label you approve, fixes a failing lint check, pushes, and reports when the PR merges. ## Quick Start Watch my current branch's CI, fix any failures that come up, and keep pushing until the PR is green and merged.

Frequently Asked Questions about babysit

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

FAQPage Schema
How do I automatically fix failing CI checks on a GitHub PR?

Use a watch loop that polls check status with gh pr checks, pulls failing job logs with gh run view --log-failed, reproduces the failure locally using the exact workflow command, then commits and pushes the fix. The loop re-arms against the new head SHA after each push.

How to monitor CI on Forgejo or Gitea pull requests?

Use the tea CLI: query commit status with tea api on the commits endpoint, list runs with tea actions runs list, and fetch job logs with tea actions runs logs. The watch, classify, and fix loop is identical to the GitHub flow.

Does this work on a branch without an open PR?

Yes. Without a PR, the loop reads the commit status of the head SHA directly via the check-runs and status APIs, using the remote's default branch or an explicitly named base. The branch must already be pushed.

What happens when CI keeps failing due to infrastructure issues?

Infra failures such as runner errors or network timeouts get a re-run via gh run rerun --failed after a 10-15 minute cooldown. After 3 strikes per check, the loop stops and reports the exact action a human must take.

When should I not use an automated CI babysitting loop?

Do not use it to create the commit series or PR itself, to watch many PRs as an inbox, or when a fix would change product behavior, public APIs, or schemas. Those decisions require explicit user direction.