skill-stocktake

You can audit Claude skills for quality assurance by running an automated evaluation against a defined checklist and AI judgment. This process identifies outdated, overlapping, or low-quality commands, providing detailed reports on skill status, reasons, and potential improvements or retirement.

A quick scan evaluates recently changed AI skills, while a full stocktake performs a comprehensive review of all global and project-level commands. Both modes use sequential subagent batch evaluation to generate detailed reports on skill status and potential improvements.

You perform an automated skill audit by triggering a sequential subagent batch evaluation against your command set. The system checks global and project-level skills against a defined checklist, evaluating quality and generating actionable improvement or retirement reports.

Run a quick scan to target only recently changed skills for fast quality assurance checks. Use a comprehensive full stocktake when you need to evaluate the entire command set across global and project levels to identify overlapping or low-quality skills.

Yes, the skill audit process explicitly supports both global and project-level commands. It evaluates all targeted commands using AI judgment and a checklist, ensuring your entire Claude environment maintains a clean and efficient command set.