skill-creator

To create a GLM Code skill, you scaffold a SKILL.md file with a name and description, then run an automated workflow to validate the required frontmatter structure and standardize the drafting process.

The best way to improve skill triggering accuracy is to run an evaluation loop that collects trigger evaluation outcomes and uses those results to guide automated description optimization and content refinement.

Iterative refinement works by orchestrating trigger evaluations, logging historical results, and applying those outcomes to enhance skill content and descriptions until stable performance is achieved.

Yes, you need yaml configured in your environment, as the skill creation workflow depends on it to validate frontmatter and orchestrate scaffolding, evaluation, and packaging utilities.

You package a validated skill by running built-in validation utilities to confirm readiness, which then prepares the skill content into a distributable format for deployment.

A skill description fails to trigger correctly when it lacks optimization, which you can resolve by iterating through the evaluation loop and applying the automated description optimization based on logged results.