heal-ci

You can automatically classify CI failures by analyzing failure logs to categorize them as flakes, defects, or unknowns. The classification routes flakes to rerun once and defects to file a report, reducing manual intervention in CI/CD pipelines.

You need to triage CI failures with GitHub Actions when a CI run fails and you want to automatically decide whether to rerun the failed job once or file a defect report for further investigation instead of manually reviewing logs.

Yes, automated CI failure routing requires GitHub integration. The Skill uses GitHub Actions and API calls to analyze failure logs and route known transients to rerun or file defects for triage within your existing workflow.

The best way to handle flaky CI jobs is to automatically classify them as flakes, defects, or unknowns based on logs. Flakes are routed to rerun once, while defects are filed as reports, speeding up resolution without manual intervention.

You trigger automated CI failure analysis by using the command /heal-ci or heal CI for #N when a CI run fails. The Skill then analyzes the logs and decides whether to rerun the job or file a defect report.

The limitation of classifying CI failures is that unknown failures cannot be automatically routed to rerun or report. Only known transients are rerun once, while other failures require a defect report for further manual investigation.