What it does
Uses the GitHub CLI (gh) to find the failing checks on a pull request, fetch the GitHub Actions logs, and pull out the part of the log that explains the failure. It then summarizes the check name, run link and a short log snippet, proposes a fix plan, and implements it only after you approve. After the change it suggests re-running the tests and re-checking the PR. A bundled Python script handles the log fetching, including the case where the run is still in progress.
Scope
GitHub Actions only. Checks from other providers (for example Buildkite) are labeled external and only their link is reported. It needs gh installed and logged in with repo and workflow permissions.
Good for
A red CI check on your pull request when you want the cause and a focused fix plan quickly.
Medium risk: needs the GitHub CLI (gh) installed and logged in with repo and workflow permissions, and reads your pull request checks and CI logs over the network. The bundled script only runs read-only gh and git commands, but CI logs can contain secrets or private data, which then enter the AI's context. It is designed to propose a plan and edit code only after you approve; review the changes before you push. GitHub Actions only. Not tested: it needs a real GitHub login and pull request that the review environment did not have, so it was checked by reading the instructions and script only. Original skill by OpenAI (Apache-2.0), unmodified.