skill-creator

Create and optimize Claude Skills across the full skill lifecycle.

4|Updated Feb 12, 2026
One-click install
npx skills add https://github.com/rbaumier/skills --skill skill-creator-rbaumier
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/rbaumier/skills/tree/main/skill-creator
Command: npx skills add https://github.com/rbaumier/skills --skill skill-creator-rbaumier

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pyyaml, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill provides a structured, repeatable process to create, evaluate, and optimize Claude Skills from idea to deployed tool. It helps teams capture intent, test prompts, evaluate results, and iterate efficiently.

Core Features & Use Cases

  • Capture Intent & Draft SKILL.md for new skills
  • Automated eval loop: run_eval + improve_description + history tracking
  • Benchmarking and analysis with eval-viewer, grader, and analyzer
  • Packaging and deployment: package_skill for distributable .skill files
  • Uses across the skill lifecycle: ideation, experimentation, validation, and release

Quick Start

Create a new skill directory with SKILL.md and optional resources, then run the evaluation loop to start iterating.

Frequently Asked Questions about skill-creator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create and optimize Claude Skills from a new idea to deployment?

To create and optimize Claude Skills, you draft a SKILL.md file, run an automated evaluation loop, iterate on the results, and package the final version. This enforces frontmatter validation and progressive disclosure for safe deployments.

What is the automated evaluation loop for refining skill prompts?

The automated evaluation loop runs evaluations, improves descriptions based on results, and tracks history. It uses an eval-viewer, grader, and analyzer to benchmark prompts and ensure repeatable skill performance.

How do I package a Claude Skill for distribution?

You package a Claude Skill by using the packaging tool to bundle the SKILL.md and optional resources into a distributable .skill file. This ensures the skill is ready for repeatable and safe deployments.

Do I need pyyaml to use the skill creation and evaluation workflow?

Yes, pyyaml is a required dependency for this workflow. You need it installed in your environment to handle frontmatter validation and configuration parsing during skill creation and iteration.

What is progressive disclosure in Claude Skills?

Progressive disclosure is a design principle enforced during skill creation that structures information delivery. It ensures the SKILL.md provides necessary context incrementally, preventing overload and maintaining repeatable, safe deployments.

Can I benchmark and analyze skill performance without manual grading?

Yes, you can benchmark and analyze skill performance automatically using the built-in grader and analyzer components. These tools evaluate results within the iteration loop to provide objective scoring and history tracking.