skill-creator

You can automate prompt engineering and evaluation by orchestrating the end-to-end skill lifecycle, coordinating drafting, running evals, and generating improvement suggestions tied to measurable metrics.

Structure YAML frontmatter with a required name and description to drive deterministic tasks, and include optional bundles like scripts, references, and assets loaded during activation.

The best way to iteratively refine and benchmark skill descriptions is by running evaluation loops that collect prompts, benchmark performance, and apply description optimization across multiple iterations.

You need Python dependencies including anthropic and pyyaml to run skill evaluation workflows, which support prompt interactions and YAML frontmatter parsing for deterministic task execution.

Bundled scripts and references support deterministic skill tasks by providing executable logic and contextual data loaded as needed during activation, ensuring repeatable and reliable skill execution.

Limitations of automating skill lifecycle management include reliance on structured YAML frontmatter for activation and dependency on anthropic and pyyaml packages for executing evaluation workflows.