github-actions-actor-reliability-audit_skill

Overview

This skill audits GitHub Actions run reliability by actor to surface high-risk contributors and flaky automation owners. It ranks actors (users or bots) by failure rate, failed-run volume, and current failure streaks to help prioritize CI triage and policy decisions. Outputs are available in readable text or machine-friendly JSON for integration into pipelines.

How this skill works

The skill ingests exported GitHub Actions run JSON files and groups runs by actor or by actor+workflow, applying optional regex filters for actors, workflows, branches, events, and repositories. It calculates metrics per group: failure rate, number of failed runs, and the latest failure streak, then assigns a severity label (ok, warn, critical) using configurable thresholds. The tool prints a ranked report and can exit nonzero when configured to fail on critical groups for automated gating.

When to use it

• Identify which humans or bots are causing the most unreliable CI runs
• Enforce CI quality gates by failing pipelines when critical actors are detected
• Prioritize debugging and remediation for workflows or actors with high failure volume
• Audit historical run exports to spot flaky automation owners before they impact release
• Monitor the effect of changes to workflows or bots over time

Best practices

• Collect run JSON using gh run view into a consistent artifact directory before auditing
• Set MIN_RUNS to ignore low-volume actors and reduce noise in results
• Tune WARN/CRITICAL thresholds to match team expectations and CI scale
• Use ACTOR_MATCH / ACTOR_EXCLUDE to focus on service accounts or exclude known noisy bots
• Run in JSON mode and enable FAIL_ON_CRITICAL in CI gates for automated enforcement

Example use cases

• Run a pre-release audit to block promotion if any actor-owned workflows are critical
• Scan archived run exports to find maintainers of flaky automation for targeted remediation
• Integrate with nightly jobs to produce JSON reports consumed by dashboards or incident trackers
• Filter by repository or branch to evaluate reliability impact of a new integration or bot rollout