skill-creator

To create an AI skill, you define the skill intent, address edge cases, and write a `SKILL.md` file containing YAML frontmatter and markdown instructions, which establishes the core logic and bundles necessary resources.

You can evaluate LLM skills by running automated tests against defined test cases and comparing results to baselines. This benchmarking provides detailed analysis to facilitate iterative rewriting and performance optimization.

The best way to improve AI skill triggering accuracy is through description optimization. This feature uses an iterative loop to automatically refine skill descriptions, ensuring the skill activates precisely when intended.

Yes, you need to define test cases to benchmark prompt engineering results effectively. The skill runs these cases against your implementation, comparing outputs to baselines to provide detailed performance analysis and evaluation metrics.

Yes, you can bundle external scripts and references when building AI tools. The skill creation process supports packaging these resources alongside the `SKILL.md` file to ensure the AI has all necessary context and capabilities.