Guide creation and iterative improvement of Claude Skills with SKILL.md and evals.
npx skills add https://github.com/mrclrchtr/skills --skill skill-creator-mrclrchtr
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/mrclrchtr/skills/tree/main/.agents/skills/skill-creator Command: npx skills add https://github.com/mrclrchtr/skills --skill skill-creator-mrclrchtr
Creating and refining Claude Skills is a multi-step, error-prone process. This Skill guides users through the end-to-end workflow from capturing intent to writing the skill's frontmatter, setting up evaluations, and iterating based on results to improve triggering accuracy and reliability.
Draft a new skill by describing its goal, fill SKILL.md with the required frontmatter and body, run the evaluation loop, and iterate until triggering becomes reliable.
High-intent search queries and answers about installing and using this skill.
Claude Skill creation involves capturing user intent, drafting a SKILL.md file with required frontmatter, bundling optional resources, and iterating based on evaluations to improve triggering accuracy and reliability.
Evaluation-driven iteration is a process of generating and running evaluation prompts, collecting results, and using feedback to rewrite skill descriptions and instructions. This mechanism improves trigger accuracy and reliability for Claude Skills.
You validate a SKILL.md file by using helper scripts to parse the frontmatter and body, checking for required name and description fields, and running evaluations to ensure the skill triggers accurately before packaging for distribution.
Yes, the skill-creator workflow supports both local development and publishing workflows. It includes reusable tooling and helper scripts for packaging standalone skills or Claude plugins for distribution.
Yes, you need Python and the pyyaml dependency to use the reusable tooling. The included helper scripts rely on pyyaml to parse SKILL.md frontmatter and validate skill definitions.
A Claude Skill may not trigger reliably due to an inaccurate description or poorly defined instructions. You fix this by running iterative benchmarks, collecting evaluation results, and rewriting the skill description to improve triggering accuracy.
vuanhtu1993Create, validate, and package Claude Skills into .skill files.
BenjaminGCreate, validate, and package Claude Skills with standardized templates and scripts.
eddie-292Create, test, and iterate on Claude skills with evaluation workflows.
przbaduCreate self-contained Claude skill units with SKILL.md frontmatter and .skill packaging.
