skill-audit

To audit AI prompt quality, evaluate individual skills against a 12-category rubric to identify behavioral gaps and ensure reliable performance across workflows. This process uses deterministic, evidence-based verification to enforce completion.

A convergence loop in skill evaluation is an iterative verification mechanism used to enforce phase gating and ensure behavioral quality gaps are fully resolved. It is needed when evaluating single or batch skills for reliable performance.

Run a batch audit across multiple AI skills by executing a multi-mode review that detects cross-skill patterns using a 12-category rubric. This surfaces behavioral quality gaps and records decisions to a persistent state file.

Yes, you can save audit decisions to a persistent state file for auditability. The interactive, category-by-category evaluation records implemented decisions directly to the state, ensuring reliable performance tracking.

The 12-category rubric for self-audits is limited to evaluating individual skills, AI prompts, and integration touchpoints. It enforces deterministic, evidence-based verification and does not assess factors outside its defined convergence loops.