skill-creator

To create Claude-based skills, draft an initial SKILL.md, run trigger-evaluation prompts on representative queries, review results, and iterate until triggering is reliable. This guides the end-to-end process from concept to deployment.

Evaluating skill triggering accuracy involves running eval prompts on a representative set of queries to measure discovery and optimize descriptions. Iterative improvement loops refine the triggering reliability.

Prompt engineering improves skill discovery by optimizing triggering descriptions based on eval results. Structuring the SKILL.md and iterating on prompt design ensures the skill triggers reliably for automation workflows.

Yes, building and iterating on skills requires the anthropic and pyyaml dependencies to structure SKILL.md files, run evaluations, and process YAML configurations for documentation and automation.

A SKILL.md file structures Claude-based skills by capturing best practices, defining triggering descriptions, and documenting creation processes for scalable deployment across domains.

Unreliable skill triggering occurs when descriptions are not optimized. Running trigger-evaluation prompts on representative queries and performing iterative improvement loops will measure accuracy and fix discovery issues.