system-skill-creator

To create and evaluate LLM skills, draft the SKILL.md frontmatter and instructions, generate test cases, and run paired with-skill and baseline evaluations to measure and improve triggering accuracy.

Benchmarking LLM skill behavior requires generating organized test cases, executing paired with-skill and baseline runs, and capturing timing and grading metadata to aggregate quantitative evaluation results.

Optimize skill descriptions for higher invocation rates by applying description-optimization loops that use quantitative assertions and user feedback to iteratively refine skill behavior and triggering.

Yes, you can iterate on skill drafts by applying quantitative assertions and user feedback within optimization loops to refine skill behavior, evaluate outputs, and improve overall triggering accuracy.

Yes, organized test cases are required to run skill evaluations, serving as the foundation for paired with-skill and baseline runs, capturing timing and grading metadata, and generating review artifacts.

LLM skills fail to trigger reliably when descriptions lack optimization; applying systematic evaluation, paired baseline runs, and description-optimization loops resolves triggering inaccuracies in developer workflows.