resolve-ci-failures

To debug GitHub Actions failures, this Skill fetches failed job logs using the gh CLI, categorizes the root cause as lint, type, test, build, or infrastructure issues, and proposes targeted fixes for approval before rerunning the pipeline.

You can automatically rerun CI jobs after fixes by having the Skill run full-project checks, issue a critique, and monitor the subsequent GitHub Actions run to confirm the pipeline succeeds without requiring manual triggers.

The best way to categorize CI failures is to parse the failed job logs and classify issues into deterministic types like lint, type, test, or build errors for automated fixes, while flagging unstable infrastructure problems for human intervention.

This Skill does not require human approval for all fixes; it automatically resolves deterministic lint, type, test, and build errors, but explicitly requires human approval for non-infrastructure fixes and unstable environments.

Using the GitHub CLI to investigate failing CI/CD workflows allows the Skill to execute commands like 'gh run view --log-failed' to fetch exact failure logs, enabling accurate diagnosis and coordinated remediation tied to the current branch context.

When CI regressions block deployment, you should gather the current branch's failing jobs, summarize the categorized failures, assign focused agent tasks to apply approved fixes, and monitor the next CI run to verify success.