skill-creator

Create Claude skills with SKILL.md frontmatter and evaluation loops.

1|Updated Feb 26, 2026
One-click install
npx skills add https://github.com/supersterling/primus --skill skill-creator-supersterling
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/supersterling/primus/tree/main/.agents/skills/skill-creator
Command: npx skills add https://github.com/supersterling/primus --skill skill-creator-supersterling

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires anthropic, pyyaml.

What problem does it solve?

This Skill unit provides a repeatable workflow to design, test, and iterate Claude skills. It guides you from drafting SKILL.md through evaluation and refinement, enabling teams to ship reliable triggers and scalable skill ecosystems.

Core Features & Use Cases

  • Structured frontmatter and clear instructions to create new skills from scratch.
  • Built-in patterns for evaluating prompts, collecting feedback, and iterating for better triggering accuracy.
  • Support for bundling resources (scripts, references, assets) and packaging into deployable skill files.
  • Use cases include creating new skills, updating existing ones, running eval loops, benchmarking, and optimization of descriptions for better triggering.

Quick Start

Draft a new skill by creating SKILL.md with descriptive frontmatter, then run the eval loop to iterate and improve.

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 evaluate Claude skills for better triggering accuracy?

To create Claude skills, draft a SKILL.md with structured frontmatter and clear instructions, then run an evaluation loop to iterate and improve triggering accuracy and reliability.

What is the best way to structure frontmatter for a new Claude skill?

The best way to structure frontmatter for a Claude skill is to define descriptive metadata within SKILL.md, which provides a repeatable blueprint for designing, testing, and shipping scalable skill ecosystems.

Can I bundle scripts and references when building Claude skills?

Yes, you can bundle optional scripts, references, and assets when building Claude skills, packaging them into deployable skill files for comprehensive automation workflows.

Do I need Python and pyyaml to iterate through skill evaluation loops?

You need Python with the pyyaml and anthropic dependencies to configure and run the evaluation loops required for benchmarking and optimizing Claude skill descriptions.

Why does my Claude skill trigger unreliably and how can I fix it?

Claude skills trigger unreliably due to poorly optimized descriptions; fix this by running built-in evaluation loops to collect feedback, benchmark performance, and iterate on the SKILL.md body.