skill-optimizer

To reduce token usage in SKILL.md, apply layered loading by keeping core sections concise and moving extended content to external references. This optimization shrinks prompt size while preserving the skill's core functionality.

Auditing skill repositories for consistency requires analyzing SKILL.md frontmatter and content across the entire repository. This automated process identifies unnecessary token usage and optimizes prompts for efficiency.

Yes, skill optimization requires a well-formed SKILL.md with frontmatter containing name, description, triggers, and keywords. A concise core section is also needed, with optional extended content linked via external references or assets.

To move large templates and examples out of SKILL.md, use external references. Relocating these assets maintains a lean core SKILL.md file, deferring extended content loading until it is actually needed.

Avoid layered loading when a SKILL.md lacks well-formed frontmatter with triggers and keywords. Without a concise core section to anchor the skill, deferring extended content to external references may break core functionality.