skill-creator

Draft, test, and iterate Claude Code skills with SKILL.md and eval loops.

27|4|Updated Mar 6, 2026

Search Tags:#automation #evaluation #iteration #benchmark #skill-development #claude-code

Ready to run

npx skills add https://github.com/skateddu/claude-code-python-setup --skill skill-creator-skateddu

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/skateddu/claude-code-python-setup/tree/main/.claude/skills/skill-creator
Command: npx skills add https://github.com/skateddu/claude-code-python-setup --skill skill-creator-skateddu

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill helps users create new Claude Skills, modify and improve existing ones, and measure skill performance. It covers drafting skills, running evals, benchmarking, and optimizing trigger descriptions to increase triggering accuracy.

Core Features & Use Cases

Capture Intent: Understand what the user wants to achieve and translate it into a Skill.
Interview and Research: Gather edge cases, dependencies, and test scenarios to shape the Skill.
Write the SKILL.md: Produce a clean frontmatter and a detailed body of instructions.
Progressive Disclosure: Use metadata, SKILL.md body, and bundled resources in layers for fast loading.
Run Evaluations: Execute structured prompts, compare with baselines, and iterate to improve performance.
Expand & Package: Add scripts, references, and assets; validate and publish as needed.

Quick Start

Draft a new capability, run an evaluation loop, and iteratively improve the description and the skill based on results.

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 test new Claude Code skills?▼

To create and test new Claude Code skills, you draft a SKILL.md file with frontmatter and instructions, run structured evaluation loops against baselines, and iteratively refine the trigger descriptions to optimize performance.

What is the best way to improve Claude skill triggering accuracy?▼

Improving Claude skill triggering accuracy involves running structured evaluations, comparing outputs against baselines, and iteratively optimizing the trigger descriptions in the SKILL.md metadata based on benchmark results.

How does progressive disclosure work when building Claude skills?▼

Progressive disclosure in Claude skills uses layered resources like metadata, the SKILL.md body, and bundled scripts or assets to support deterministic tasks while maintaining fast loading times.

Can I use Python scripts and YAML dependencies to package Claude skills?▼

Yes, you can package Claude skills using a reusable resource structure that includes scripts, references, and assets, relying on Anthropic and YAML dependencies to support deterministic task execution.

How do I benchmark Claude skills to measure performance?▼

You benchmark Claude skills by executing structured prompts, comparing the results with established baselines, and iterating on the skill instructions and descriptions to achieve measurable performance improvements.

What should I do if my Claude skill is not triggering correctly?▼

If your Claude skill is not triggering correctly, you should run evaluation loops to test structured prompts, analyze the benchmark results, and iteratively optimize the trigger descriptions in your SKILL.md frontmatter.