skill-creator

Automate prompt engineering and skill evaluation by running repeatable test loops that benchmark triggering accuracy and performance. The tool auto-generates skills from prompts, validates them through eval cycles, and iteratively refines them to reduce guesswork.

Benchmark triggering accuracy by running automated evaluation cycles that test skills against dynamic prompts. This process measures performance, compares iterations, and manages history to drive data-driven improvements for optimized skill deployment.

Design and validate skills by pointing the tool at a directory containing SKILL.md and supporting assets to run a stabilization loop. This automates end-to-end refinement, validating frontmatter-driven discovery and dynamic prompts through repeatable testing.

Yes, you can package modular skills for distribution after benchmarking. The tool compares different iterations, evaluates performance metrics, and packages the optimized skills for deployment to ensure reliable triggering accuracy.

Yes, the skill evaluation tool supports frontmatter-driven discovery and dynamic prompts. It uses these features during automated benchmarking to optimize triggering accuracy and guide users through iterative improvements.

Skills require iterative improvements because initial auto-generation needs validation through repeatable test loops. Running eval cycles benchmarks performance, identifies triggering inaccuracies, and manages history to drive necessary data-driven refinements.