ci-fix-monitor

Diagnose CI check failures on pull requests and push fixes until green.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/pandejesal/drone-nav-sar --skill ci-fix-monitor-pandejesal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ci-fix-monitor
Source: https://github.com/pandejesal/drone-nav-sar/tree/main/.swarm/bundled-skills/ci-fix-monitor
Command: npx skills add https://github.com/pandejesal/drone-nav-sar --skill ci-fix-monitor-pandejesal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Red CI checks on a pull request block merging, and manually reading logs, classifying failure types, and iterating on fixes is slow and error-prone. This Skill automates the full loop: fetch check status, classify each failure, apply the correct fix, push, and re-monitor until all checks pass. ## Core Features & Use Cases - Failure classification: Distinguishes check-title, package-check, branch-drift, lint/format, unit test, integration, macOS file I/O, security, and smoke failures, each with a targeted fix path. - Log-driven diagnosis: Fetches failing job logs via GitHub CLI or GitHub MCP tools, and separates PR-introduced failures from pre-existing failures on main. - Bounded fix-push loop: Rebases onto main for version drift, applies Biome format/lint fixes only to changed files, and caps standalone retries at 5 cycles before escalating. - Use Case: A PR shows red checks for a Biome format violation and a stale branch behind main. The Skill rebases onto origin/main, force-pushes with --force-with-lease, auto-formats the changed files, and waits for the next CI event instead of polling. ## Quick Start Monitor the CI checks on my open pull request, diagnose any failures, fix them, and re-push until everything is green.

Frequently Asked Questions about ci-fix-monitor

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 pull request?

Fetch the PR's check runs, read the failing job logs, classify each failure by type (title, package-check, lint, test, drift), then apply the matching fix and push. Repeat the cycle until all required checks pass, capping retries at five iterations.

How do I fix a branch behind main CI failure?

A branch-behind-main failure happens when main received a release commit after your branch was cut. Run git fetch origin main, git rebase origin/main, then git push --force-with-lease, since a regular push is rejected after rebasing.

Why do unit tests fail on macOS CI but pass on Linux and Windows?

macOS/APFS has different filesystem timing, so an atomic rename can complete before data is visible to subsequent reads. Fix write-then-read races with a bunWrite plus ENOENT retry pattern and Node FileHandle.sync() instead of fsync.

Can I use GitHub MCP tools instead of the gh CLI for CI monitoring?

Yes, GitHub MCP tools replace gh commands in remote environments: use pull request read with get_check_runs for status, get_job_logs for log content, and update pull request for title edits. Resolve actual tool names by capability since MCP names vary across environments.

When should I stop retrying CI fixes and escalate?

Stop after five fix-push cycles if the PR is still not green. Escalate to the user with the last failing check name and a short log excerpt rather than looping indefinitely, and never declare success until all required checks pass.