skill-creator

You can automate Claude skill creation by drafting a skill and running a structured evaluation loop to benchmark trigger performance and refine prompts until quality improves. This orchestrates the end-to-end development workflow automatically.

Benchmarking trigger performance involves testing configurations across an evaluation loop and tracking progressive improvements. This approach measures triggering accuracy and refines descriptions to ensure skills activate reliably in workflows.

Iterating prompts requires running initial evaluations, analyzing benchmark results, and refining tests in a repeatable loop. This structured iteration turns rough drafts into robust, trigger-ready capabilities by optimizing descriptions and prompts.

Yes, you need the pyyaml dependency installed to run the evaluation scripts. The skill creation workflow relies on this package to parse configurations and manage the automated benchmarking components.

You can optimize skill descriptions automatically by generating and refining them through the built-in evaluation loop. This process maximizes triggering accuracy by benchmarking description variations and selecting the most effective configuration.

Skills fail to trigger correctly when descriptions lack optimization or benchmarking. Running the skill through the evaluation loop identifies trigger performance issues and refines the prompt to resolve activation failures.