skill-creator

To create and test custom AI skills, you draft instructions, define trigger conditions, and shape evaluation-ready metadata into a tested package. This iterative workflow supports test prompt design, benchmark comparison, and grading to refine performance.

Skill benchmarking is the process of comparing outputs with and without a skill to track quantitative performance and inspect qualitative results. It improves prompt iteration by using grading feedback to optimize description accuracy and workflow quality.

The best way to design evaluation-ready skill metadata is to satisfy frontmatter requirements while shaping operational guidance and evaluation setups. This ensures reliable triggering and prepares the skill package for performance review and benchmarking.

Yes, you can use this workflow to refine trigger conditions for existing skills. It supports iterative skill development by allowing you to write new test prompts, compare benchmark results, and update operational guidance based on grading feedback.

No, you do not need prior test cases to start iterative skill development. The process includes designing test prompts from scratch, comparing with and without the skill, and refining the draft until it performs well through benchmarking and review.