ck:skill-creator

Automate Claude skill creation and benchmark trigger accuracy by using eval-driven iteration scripts that run parallel with-and-without-skill tests, aggregate results, and validate performance quantitatively.

Eval-driven iteration for skill engineering is a process of running parallel benchmark tests, grading outputs with analyzer agents, and aggregating data to quantitatively validate and improve trigger accuracy and overall skill performance.

Package and optimize SKILL.md descriptions by running an automated description improvement loop that iteratively refines trigger conditions, followed by quick validation scripts and packaging guidance for secure distribution.

Yes, you can scaffold SKILL.md, scripts, and references without manual trial-and-error by using initialization templates that generate the required directory structure and bundled resources automatically.

Yes, the anthropic dependency is required to run parallel eval workflows, utilize grader and comparator agents, and execute the automated description optimization loop for iterative skill improvement.

The best way to validate skill performance before distribution is to run the integrated eval suite for quantitative benchmarking, analyze post-hoc aggregation reports, and execute quick validation packaging checks.