common-skill-creator

To create token-efficient AI agent skills, use compact frontmatter-first templates and enforce a three-level loading system to maximize context relevance. This minimizes unnecessary token consumption while maintaining instruction quality.

The best way to evaluate AI agent skills is by implementing a formal eval workflow using evals.json. This process involves running should-trigger queries and regression checks to verify trigger accuracy and skill quality.

Organize references, scripts, and assets for AI agent skills by strictly applying a three-level loading system. Proper resource placement ensures efficient loading and retrieves necessary context only when required by the agent.

Yes, this skill creation framework works with Cursor and Copilot. The standards apply to creating, testing, and optimizing high-density Agent Skills for any AI Agent, including Claude, Cursor, Windsurf, and Copilot environments.

Audit and compress existing AI agent skills by applying frontmatter discipline and a three-level loading system. This process improves trigger accuracy and reduces total token usage without losing functional context.

You need evals.json for your agent skill workflow to drive quality through formal regression checks. It defines should-trigger queries and token-budget considerations, ensuring your skill activates accurately and efficiently.