What problem does it solve?
Prevents fragile AI-generated changes from reaching human review by automatically running tests, checking spec coverage, and validating lint/build signals first.
Core Features & Use Cases
- Automated verification gate between apply and review: Selects a change and the next completed Task Group eligible for verification, then runs evidence-gathering checks without requiring human input.
- Best-effort test execution: Detects common test infrastructures (e.g., pytest, npm, bun, Go) and records full output; failures are captured as warnings rather than blocking logic except where build verification is critical.
- Spec coverage assessment: Locates group specs under openspec/changes/<name>/specs/<capability>/spec.md and classifies each requirement as Full, Partial, or Missing based on implementation evidence.
- Lint/build verification: Detects common linters and build commands; captures results to support the review decision.
- Structured Verify Report + tracking-safe posting: Generates a consistent report and posts only a note to existing child issues when GitHub/GitLab tracking metadata is present.
Quick Start
Run verify for your next completed Task Group by asking the AI to execute the corgispec-verify skill for the currently selected change.