skills.rest
OiAnthony
OiAnthonyCOMMUNITY
Explore all skills by this author →
GitHub

skill-creator

Create, test, and iteratively improve Claude skills from SKILL.md files.

8|Updated Jan 19, 2026
Search Tags:#automation#claude code#benchmarking#skills#skill writing#evals#trigger-optimization
One-click install
npx skills add https://github.com/OiAnthony/.agents --skill skill-creator-oianthony
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/OiAnthony/.agents/tree/main/skills/skill-creator
Command: npx skills add https://github.com/OiAnthony/.agents --skill skill-creator-oianthony

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

It helps you design, validate, and improve Claude skills without guessing which instructions will trigger reliably or which workflows need stronger guardrails.

Core Features & Use Cases

  • Draft or rewrite SKILL.md frontmatter and body content for a new or existing skill.
  • Generate realistic trigger evals, split them into train and holdout sets, and compare with-skill versus baseline behavior.
  • Grade outputs, surface flaky assertions, aggregate benchmark statistics, and package a finished skill for distribution.
  • Use it when you want to turn a rough idea into a tested skill, tighten skill descriptions for better triggering, or diagnose why a skill underperforms.

Quick Start

Use the skill-creator skill to turn your rough idea into a tested SKILL.md, run realistic evals against it, and iterate until the description triggers reliably.

Frequently Asked Questions about skill-creator

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

FAQPage Schema
How do I test and benchmark Claude skills against a baseline?▼

To benchmark Claude skills, generate realistic trigger evals, split them into train and holdout sets, and run parallel with-skill versus without-skill executions to grade outputs and aggregate statistics.

How do I improve trigger accuracy for a SKILL.md file?▼

Improve SKILL.md trigger accuracy by iteratively rewriting frontmatter and body content, then applying description optimization scripts to test and refine how reliably the skill triggers against prompts.

What is the best way to create a Claude skill from a rough idea?▼

Create a Claude skill by drafting a single SKILL.md source of truth, generating an evaluation set, and iteratively validating behavior against realistic prompts to tighten instructions and guardrails.

Do I need PyYAML and Claude to write and evaluate skills?▼

Yes, writing and evaluating skills requires Claude to run parallel without-skill baseline tests and PyYAML to process SKILL.md frontmatter during drafting, benchmarking, and description optimization.

Why does my Claude skill underperform or fail to trigger reliably?▼

A Claude skill underperforms when frontmatter descriptions lack optimization or guardrails are weak; diagnose this by grading outputs, surfacing flaky assertions, and comparing baseline behavior with-skill.

Can I package and distribute a skill after benchmarking it?▼

Yes, after grading outputs, surfacing flaky assertions, and aggregating benchmark statistics, you can package the finished skill for distribution directly from the validated SKILL.md source.

Related Skills

Explore More Software Engineering Skills→
softee-psoftee-p

skill-creator

Compile user ideas into reproducible Claude Skill units with SKILL.md and evaluations.

Community
Advanced
gelowzzzgelowzzz

skill-creator

Create, validate, and package Claude Skills into .skill files.

Community
Advanced
gesmith0606gesmith0606

skill-creator

Create, evaluate, and iteratively improve SKILL.md descriptions with Python scripts.

Community
Advanced
iadr-deviadr-dev
2

skill-creator

Create and iteratively improve Claude Code skills with test-driven evaluation.

Community
Advanced