skill-creator

You create and test AI skills by defining intent, drafting a `SKILL.md` file, setting up test cases, and running evaluations. This iterative development process uses feedback loops and performance benchmarks to refine skill logic and optimize descriptions for accurate triggering.

The best way to optimize AI skill descriptions is by testing the `description` field against established test cases. Running evaluations and applying performance benchmarks allows you to iteratively refine the description to improve triggering accuracy.

Iterative development for prompt engineering works by running evaluations on test cases, collecting performance metrics, and refining the `SKILL.md` logic. Feedback loops guide the optimization process until the AI skill meets the desired performance benchmarks.

Yes, you can evaluate AI skill performance using included scripts and references to set up test cases. The skill-creator guides you through running these evaluations and measuring performance benchmarks without requiring extensive external custom scripts.

Yes, you need to structure skill directories specifically when building AI skills. Structuring directories correctly ensures that the `SKILL.md` file, scripts, references, and assets are properly organized for the skill to function and trigger accurately.

AI skills fail to trigger accurately in evaluations when the `description` field is poorly optimized. You must run iterative development tests, analyze performance benchmarks, and refine the description to improve the skill's triggering accuracy.