Create and refine Claude Skills with SKILL.md and trigger-eval loops.
npx skills add https://github.com/lDuang/mcp-skill-registry --skill skill-creator-lduang
Please help me install this Agent Skill. Skill: skill-creator Source: https://github.com/lDuang/mcp-skill-registry/tree/main/skill/skill-creator Command: npx skills add https://github.com/lDuang/mcp-skill-registry --skill skill-creator-lduang
Provides a structured, repeatable workflow for creating new Claude Skills, iterating on their SKILL.md frontmatter and body, running trigger-eval loops, and benchmarking results to drive improvements. It coordinates drafting, evaluation, feedback collection, and packaging into distributable skill artifacts, helping teams scale skill creation with quality guardrails.
Start the iteration loop with your eval set and skill directory to begin optimizing the skill trigger description.
High-intent search queries and answers about installing and using this skill.
To create and iterate on Claude skills, draft SKILL.md files with YAML frontmatter, run trigger-eval loops, collect benchmark results, and refine descriptions. The Skill Creator orchestrates this end-to-end lifecycle for rapid, repeatable improvements across multiple iterations.
A trigger-eval loop is an evaluation workflow that tests skill descriptions against benchmark datasets, collects performance results, and iterates on SKILL.md content. This mechanism drives repeatable improvements in triggering accuracy and overall skill quality.
Yes, YAML frontmatter in SKILL.md is required to run skill benchmarking workflows. The Skill Creator depends on this structured metadata to coordinate drafting, evaluation, feedback collection, and packaging into distributable skill artifacts.
Yes, you can package validated Claude skills into distributable .skill artifacts after evaluation. The Skill Creator includes packaging and validation utilities that generate deployable skill files once trigger-eval loops and regression checks are complete.
You can run regression checks and parallel evaluation by configuring holdout and test splits within the evaluation workflow. The Skill Creator supports these validation workflows to ensure skill triggering accuracy does not degrade across iterations.
The best way to automate SKILL.md generation and updates is to use an orchestrated evaluation loop that drafts, evaluates, and refines file content. The Skill Creator automates this lifecycle, enabling rapid, repeatable improvements with quality guardrails.
coreyjaDraft SKILL.md frontmatter and structure guidance for Claude Code skills.
DariusCorvusDetect deployment platforms and write CLAUDE.md Deploy Configuration for gstack projects.
silvainfmCreate and manage Claude Code skills with triggers and enforcement levels.
garochee33Draft SKILL.md content, generate eval sets, and run benchmark loops.
