What problem does it solve? Agent skills often waste context tokens, violate progressive-disclosure principles, or fail to trigger because of weak descriptions, duplicate names, or runtime misconfigurations. This Skill audits every skill directory and reports these problems before they degrade agent behavior. ## Core Features & Use Cases - Token and size auditing: Measures SKILL.md character counts against a 20000-character budget and flags oversized files. - Progressive-disclosure analysis: Detects oversized sections and suggests extraction into references/ files. - Triggering and loading diagnostics: Checks descriptions for concrete trigger phrases, detects duplicate names and near-duplicate descriptions across skills, and warns about missing files, non-executable scripts, and backslash paths. - Use Case: A team maintains a dozen agent skills and notices one never triggers. Run the audit in --all mode to get a JSON report showing the description lacks trigger phrases and collides with another skill's name. ## Quick Start Ask the agent to audit all skills in .agents/skills for token waste and triggering problems using the skill-optimizer.
