skill-optimizer

Agent skills fail to trigger correctly due to weak descriptions, missing triggers, or workflow gaps in the skill metadata. Auditing skill definitions against session evidence identifies undertriggering or overtriggering issues and provides actionable optimization guidance.

To audit agent skills, analyze skill metadata, session evidence, and instruction structure to evaluate trigger quality, workflow completeness, static structure, and token efficiency. This generates a prioritized improvement report for fixing routing issues.

Yes, skill auditing applies to both Claude Code and Codex skill review tasks. It evaluates trigger analysis, workflow evaluation, and static checks to detect undertriggering, overtriggering, and optimization opportunities across these platforms.

The best way to find missing triggers is evidence-based optimization, which compares skill usage against session data. This detects routing issues caused by weak descriptions and highlights improvement opportunities without modifying skill files.

No, optimizing agent skills through evidence-based audits does not require modifying skill files. The audit analyzes metadata, session evidence, and instruction structure to provide a prioritized improvement report without altering the original files.