loop-on-ci — Claude Code skill: PR check monitoring and automated
A skill that monitors PR checks via gh pr checks and helps the agent iteratively find and fix CI failures until all tests pass. Requires GitHub CLI and repository access with an open pull request.
npx skills add cursor/plugins --skill loop-on-ci -a claude-codeWhat loop-on-ci does
loop-on-ci manages the feedback loop: polls the full set of PR checks, diagnoses why each one failed, watches pending checks with gh pr checks --watch --fail-fast, and re-evaluates the entire check suite after each commit. It uses gh pr checks as the single source of truth, which covers all PR-attached checks, not just GitHub Actions.
The skill applies fixes scoped to a single failure cause at a time, avoids bypassing hooks, and does not bloat the PR with unrelated changes. If a failure is clearly independent of the current changes and passes on main, it suggests merging the latest main instead. When tests appear flaky, it retries once and documents the evidence.
Requires gh CLI installed and permission to view and manage the PR. The skill re-runs gh pr checks --json name,bucket,state,workflow,link after every push because the check set can change.
What it helps with
- Track all PR checks including those outside GitHub Actions
- Iterate through CI failures and apply fixes until green status
- Distinguish flaky tests from real failures and retry intelligently
- Diagnose the root cause of a check failure from its logs
Example prompts for the agent
- Monitor the PR status and help fix any failing checks
- Run loop-on-ci on the feature branch and get all checks to green
- Check PR status, find what broke, and fix the code
How to install loop-on-ci
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill loop-on-ci -a claude-code2Manually: put the files in place
~/.claude/skills/loop-on-ci/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/loop-on-ci/
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill loop-on-ci -a codex2Manually: put the files in place
.agents/skills/loop-on-ci/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill loop-on-ci -a cursor2Manually: put the files in place
.cursor/skills/loop-on-ci/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install cursor/plugins loop-on-ci2Manually: put the files in place
.github/skills/loop-on-ci/SKILL.md1Skills CLI (Vercel)
npx skills add cursor/plugins --skill loop-on-ci -a gemini-cli2Platform CLI
gemini skills install https://github.com/cursor/plugins --path cursor-team-kit/skills/loop-on-ciManually: put the files in place
.devin/skills/loop-on-ci/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add cursor/plugins --skill loop-on-ci -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
From the same set18
The whole set →Similar
Questions about loop-on-ci
How do I install loop-on-ci for Claude Code?
Run npx skills add cursor/plugins --skill loop-on-ci -a claude-code. The skill will be available at ~/.claude/skills/loop-on-ci/SKILL.md.
What does loop-on-ci do?
It monitors PR checks using gh pr checks, diagnoses why checks fail, and helps the agent iteratively fix them. It re-evaluates the full check set after each push until all tests pass.
Why does it use `gh pr checks` instead of `gh run list`?
gh pr checks includes all checks attached to the PR, while gh run list only covers GitHub Actions workflows. This gives a complete picture of check status.
