skill-creator-skill

Guide Claude skill creation from intent capture through evaluation and refinement.

Updated Mar 2, 2026
One-click install
npx skills add https://github.com/dacy/architect-skill --skill skill-creator-skill
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-creator-skill
Source: https://github.com/dacy/architect-skill/tree/main/.claude/skills/skill-creator-skill
Command: npx skills add https://github.com/dacy/architect-skill --skill skill-creator-skill

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

The Skill Creator helps teams design and refine Claude skills from idea to production, providing structured guidance for capturing intent, interviewing stakeholders, and writing robust SKILL.md specifications. It supports iterative evaluation, benchmarking, and triggering-accuracy improvements to reduce undertriggering and overtriggering.

Core Features & Use Cases

  • Capture Intent: understand user goals, context, and success criteria to craft a clear skill scope.
  • Interview and Research: systematically gather edge cases, dependencies, and test inputs; consult available MCPs or docs as needed.
  • Write the SKILL.md: generate a frontmatter with name, description, compatibility, and a body that instructs how to apply the skill; include a pattern for test prompts and evaluation workflows.
  • Run Evals and Iterate: set up evaluation prompts, run trials, view results, and rewrite the skill based on feedback.
  • Optimize Triggering: after evaluation, refine the description to improve discovery and reduce misfires, using the built-in improvement script.

Quick Start

Draft a skill idea, identify key prompts and metrics, and begin the write-evaluate loop to generate a first usable SKILL.md.

Frequently Asked Questions about skill-creator-skill

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

FAQPage Schema
How do I create and test Claude skills with automated evaluation workflows?

To create and test Claude skills, you capture intent, draft SKILL.md frontmatter, design test prompts, run evaluations, and iteratively refine descriptions to maximize trigger accuracy.

What's the best way to optimize skill triggering accuracy and reduce misfires?

Optimize skill triggering by running evaluations, benchmarking results, and using the built-in improvement script to rewrite descriptions, reducing both undertriggering and overtriggering misfires.

Do I need anthropic and pyyaml dependencies to build and benchmark skills?

Yes, the skill creation workflow requires the anthropic and pyyaml dependencies to generate specifications, run evaluation trials, and benchmark trigger accuracy during iterative refinement.

How does the skill interview process capture edge cases and dependencies?

The skill interview process systematically gathers edge cases, dependencies, and test inputs by consulting available MCPs or documentation to define a robust scope and success criteria.

Can I iterate on SKILL.md specifications based on evaluation feedback?

Yes, you can iterate on SKILL.md specifications by running evaluation prompts, viewing trial results, and rewriting the skill body and frontmatter based on the evaluation feedback.

Why does my Claude skill undertrigger or overtrigger during prompt evaluation?

Skills undertrigger or overtrigger when the description lacks specificity; you resolve this by benchmarking evaluation results and refining the description to improve discovery accuracy.