auto-test-skill

Automate multi-round testing by generating structured A-round evaluation plans and B-round quality checks. This pipeline orchestrates critique and testing to ensure traceable improvements for AI skills in local repositories.

Independent A-round evaluation avoids leakage by isolating the assessment phase from the testing phase. It automatically generates separate planning and testing documents in the plans/ and tests/ directories to maintain strict process separation.

Configure YAML-driven quality thresholds by defining criteria such as min_p0_p1_ratio and min_systemic_issues. The B-round quality checks validate generated plans and tests against these configurable thresholds to ensure measurable improvements.

Yes, multi-round skill optimization supports cross-skill applicability across multiple skills in a local repository. You trigger the testing pipeline against a target skill root to automatically emit A-round plans and B-round checks.

A-round plans focus on independent evaluation and structured critique, while B-round quality checks validate those plans against configurable YAML thresholds. Together, they enforce traceable outputs and ensure measurable skill improvements.

You need YAML-driven threshold validation when enforcing measurable quality criteria during multi-round skill testing. It validates outputs against configurable metrics like min_p0_p1_ratio and min_systemic_issues to guarantee traceable improvements.