optimize-skill-description

To optimize skill descriptions and prevent trigger misfires, design evaluation queries, split train/validation data, run multi-run trigger-rate evaluations, and generate analysis reports to drive iterative improvements.

Evaluating trigger reliability involves maintaining a 60/40 train/validation split, running progressive disclosure evaluations, and using near-miss negative examples to sharpen activation boundaries.

Designing evaluation queries requires creating targeted prompts that test skill activation, splitting them into train and validation sets, and measuring trigger rates across multiple runs to identify misfires.

Wrong skill triggers occur when descriptions lack boundary clarity. Using near-miss negative examples during evaluation sharpens activation rules and reduces misfires across similar skills.

Yes, YAML frontmatter with name, description, license, and metadata at the root is required to optimize skill descriptions and run trigger-rate evaluation workflows.