ywc-toolkit-eval

To evaluate Claude Code skills and agents, use a two-tier scoring harness that performs structural compliance checks and behavioral validation. This graded scorecard approach assesses activation precision across six quality axes to maintain catalog integrity.

The best way to identify redundant skills is by running an evaluation that generates a prioritized backlog. It automatically ranks the weakest items and tracks historical performance trends to highlight drifted behavioral contracts.

Automated agent scoring drives continuous improvement by applying mechanical and judgment-based evaluations to custom agents. It identifies behavioral efficacy issues and operational drift, producing actionable feedback to optimize toolkit performance.

You should perform a quality assessment after major toolkit updates. This detects structural compliance issues and activation accuracy drift, ensuring your catalog remains lean, accurate, and free of redundant skills or agents.

Yes, grading your Claude Code toolkit minimizes token waste by identifying and removing redundant skills. The evaluation process ensures operational efficacy and structural compliance, keeping your catalog lean and efficient.