Create and iteratively refine Claude Code skills with trigger evaluations.
npx skills add https://github.com/Axeloooo/TheDataBay --skill skill-creator-axeloooo
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/Axeloooo/TheDataBay/tree/main/.agents/skills/skill-creator Command: npx skills add https://github.com/Axeloooo/TheDataBay --skill skill-creator-axeloooo
This Skill helps you design new Claude Code skills and iteratively improve existing ones so they trigger reliably and perform well on real tasks.
Tell this Skill what you want your skill to do, when it should trigger, and the expected output format, and it will guide you through drafting the Skill and running evaluation test prompts.
High-intent search queries and answers about installing and using this skill.
You improve Claude Code skill triggering accuracy by running quantitative trigger evaluations and iteratively refining the SKILL.md frontmatter description through benchmarking and variance analysis to reduce under-triggering and false triggers.
Creating a deterministic Claude Code skill draft involves defining the target capability, trigger conditions, and expected output format to generate a properly structured Skill with a strong name and description entry point for SKILL.md frontmatter discovery.
Iterative improvement with benchmarking works by repeatedly testing the skill description against generated test prompts, using variance-aware reporting to guide changes and optimize the resulting behavior for intended task outcomes.
Yes, you need PyYAML installed because the skill relies on it to support structured evaluation loops and frontmatter processing required for trigger-accuracy evaluation and iterative description refinement.
Under-triggering and false triggers occur when the SKILL.md frontmatter description is not properly optimized, requiring iterative refinement via test prompts and run_eval tooling to align the trigger conditions with intended user queries.
Yes, you can use run_eval and improve_description tooling to execute structured eval loops, apply variance-aware reporting, and iteratively refine skill drafts to ensure the resulting behavior meets intended task outcomes.
syahiidkamilGuide users through designing, documenting, and evaluating Claude Code skills.
onzhangCreate and evaluate Claude Skills using SKILL.md and Python scripts.
tomhouleDefine and validate SKILL.md frontmatter files for agent capability discovery.
xabierlameiroCreate, modify, test, and benchmark AI skills with variance analysis.
