Draft and refine Claude skill descriptions with evaluation loops and benchmarking.
npx skills add https://github.com/cogni-dao/cogni --skill skill-creator-cogni-dao
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/cogni-dao/cogni/tree/main/.agents/skills/skill-creator Command: npx skills add https://github.com/cogni-dao/cogni --skill skill-creator-cogni-dao
The Skill Creator provides a structured, repeatable workflow to design, test, and continuously improve Claude skills, reducing guesswork and misfires in activation.
Create a new skill directory with a SKILL.md frontmatter and start the evaluation loop to refine its triggering behavior.
High-intent search queries and answers about installing and using this skill.
To build and test AI skills, you draft descriptions, run evaluation loops, and benchmark versions iteratively. This structured workflow reduces misfires in activation by validating triggering behavior before deployment to production.
Optimizing Claude skill descriptions requires iterative improvements and benchmarking across versions. By running evaluation loops, you can refine frontmatter-driven discovery to maximize triggering accuracy and ensure robust performance.
Yes, PyYAML supports managing frontmatter-driven discovery and optional dependencies for skill deployment. This allows you to validate the SKILL.md file, scope each skill correctly, and package bundled resources like scripts and references.
Benchmarking AI skill performance involves using integrated evals and iteration history to objectively compare versions. This allows you to track triggering behavior improvements and ensure the skill meets robust performance standards before production deployment.
No, scripts, references, and assets are bundled resources supporting end-to-end deployment. Packaging workflows include these components to ensure reuse in future skills and seamless onboarding without separating them manually.
AI skills fail to trigger accurately due to poorly scoped descriptions and lack of validation. Implementing evaluation loops and frontmatter-driven discovery reduces guesswork, ensuring the skill is well-scoped and easy to onboard.
vuanhtu1993Create, validate, and package Claude Skills into .skill files.
coreyjaDraft SKILL.md frontmatter and structure guidance for Claude Code skills.
Speirsy11Automate drafting, evaluating, and iterating Claude skills with SKILL.md.
datorresbAutomate skill development from SKILL.md drafting through evaluation loops and packaging.
