Skill profile · Inspected
gh-fix-ci
Inspects failing PR checks with gh, fetches the GitHub Actions logs, summarizes the failure, then proposes a fix plan and implements it only after explicit approval.
What it actually does
- Uses
ghto inspect failing PR checks and fetch the GitHub Actions logs, summarizes the failure, then proposes a fix plan and implements it only after explicit approval. - Treats non-Actions CI providers (e.g. Buildkite) as explicitly out of scope — it reports the details URL rather than pretending to handle them.
When to use it — and when not to
Use it when
- A PR's GitHub Actions checks are failing and you want a triaged fix plan before anything gets changed.
Not for
- Your CI runs on a non-GitHub-Actions provider — it names this limitation itself rather than silently failing.
The caveat that matters
It references an optionalcreate-planskill that lived in the catalog's now-removed.experimentalfolder — that specific collaboration point no longer resolves.
Agent compatibility
- Codex
Confirmed means the skill’s own source or vendor states it works with that agent. Portable means it only reaches that agent through the shared SKILL.md format and a directory convention — real, but a weaker claim. An agent with no entry here isn’t known to be unsupported; it simply isn’t documented either way, so nothing is claimed. Full model on Agent Skills Compatibility.
Requirements and operational surface
Depends on
- gh CLI
- GitHub Actions
What it actually does at runtime
- Executes
ghCLI commands - Reads GitHub Actions logs over the network
- Writes code changes only after your explicit approval
Workflow fit
Failing CI → triage → fix plan (approval gate) → implementation.
Related skills
- gh-address-commentsThe companion skill for working through PR review comments once CI is green.
Cursor plugins that relate to this
A plugin is a bundle rather than a skill — it can carry several skills, plus subagents, rules, and hooks. See Cursor plugins for the distinction.
- Cursor Team KitCursor's own 18-skill kit for CI, code review, shipping, and verification, plus two subagents and two TypeScript rules — deliberately built to work without third-party service integrations.
More from openai/skills
Other skills from the same source.
- gh-address-commentsFinds the open PR for your current branch and works through its review comments with the gh CLI, checking gh auth status first rather than failing halfway through.
- playwrightDrives a real browser from the terminal for navigation, form filling, screenshots, and data extraction — CLI-first, and won't write @playwright/test files unless asked.
- security-threat-modelProduces a repository-grounded threat model — trust boundaries, assets, attacker capabilities, abuse paths, mitigations — written to a markdown file. Roughly half its description is negative triggers.
- migrate-to-codexMigrates instruction files, skills, agents, and MCP config from another agent into Codex's own files, without touching the source agent's files.
- pdfReads, creates, and reviews PDFs where layout matters, rendering pages for visual checks rather than only extracting text.
Evidence and provenance
Maintained by OpenAI (official). Primary source: openai/skills, last checked against this page’s claims on . This page summarizes and analyzes that source — it isn’t a copy of its SKILL.md or README, and installation happens at the source, not here.
Topic: Development, DevOps & Deployment.
Also covered in The Best Codex Skills, part of the Codex guide.