skill-creator

Automate the evaluation loop for Claude skills by running automated train and test cycles. This process uses built-in evaluators to assess drafted prompts, allowing you to iterate quickly until the skill achieves robust performance.

Benchmark triggering accuracy for a skill by evaluating its performance against established benchmarks. This approach optimizes the skill's descriptions to ensure better discovery and more accurate activation during relevant queries.

Create and modify Claude skills from scratch by drafting clear SKILL.md front matter and content. Place the file in a new folder, then run the skill creation workflow to generate prompts and iterate on the design.

Yes, pyyaml is required to run the skill creation and iteration workflow. This dependency supports the scripts component, enabling the structured parsing and modification of YAML front matter within your SKILL.md files.

A skill fails to trigger correctly during evaluation if its description lacks optimization for discovery. Benchmark the skill's triggering accuracy and refine its descriptions to improve activation rates during relevant queries.