What problem does it solve? Failing CI pipelines produce noisy logs and ambiguous signals, making it hard to find the first causal failure and fix it without masking real problems or prematurely claiming success. ## Core Features & Use Cases - Failure Classification: Separates actionable failures from cancelled, skipped, or external checks, then classifies them as product code, test, workflow, dependency, flaky, or infrastructure issues. - Bounded Repair with Verification: Reproduces the failure locally, applies the smallest scoped repair, adds regression tests, and validates before any remote recheck. - Approval-Gated Remote Actions: Requires explicit approval before remote reruns, workflow mutations, merges, or deployments, and never reports a local pass as hosted CI success. - Use Case: A pull request test job fails on GitHub Actions. The skill resolves the exact run and revision, extracts the failing log section, reproduces the defect locally, adds a regression test, applies the fix, and records the remote rerun as pending approval. ## Quick Start Ask the agent to triage the failing CI run on your pull request, diagnose the root cause, and propose a verified fix.
