skill-creator

Create and improve Claude skills with tests and evaluation feedback.

44|1|Updated Mar 12, 2026
One-click install
npx skills add https://github.com/IPADS-SAI/MobiClaw --skill skill-creator-ipads-sai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/IPADS-SAI/MobiClaw/tree/main/mobiclaw/skills/skill-creator
Command: npx skills add https://github.com/IPADS-SAI/MobiClaw --skill skill-creator-ipads-sai

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

Skill Creator helps you turn an idea into a working Claude skill, then improve it with real tests instead of guesswork. It is designed for building new skills, editing existing ones, measuring performance, and tuning the description that controls when the skill triggers.

Core Features & Use Cases

  • Draft and revise skills: Shape a new skill from scratch or tighten an existing one with clearer instructions and better structure.
  • Run evals and benchmarks: Create realistic test prompts, compare skill-assisted runs against baselines, and inspect pass rates, timing, and token usage.
  • Iterate from feedback: Use grader output, benchmark notes, and review comments to spot weak instructions, flaky tests, or missing edge cases.
  • Optimize triggering: Generate and review trigger queries to improve the skill description so it activates for the right user intents more often.
  • Package for reuse: Bundle a validated skill into a distributable skill file when it is ready to share.

Quick Start

Tell Claude what skill you want to build or improve, and it will help you draft the skill, create test cases, run evaluations, and refine the description until the skill performs better.

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 Claude skills with deterministic validation?

To create and test Claude skills, you draft instructions, generate realistic test cases, and run deterministic validation by comparing skill-assisted runs against baselines to inspect pass rates and timing.

What is the best way to optimize skill trigger descriptions for the right user intents?

Optimizing skill trigger descriptions involves generating and reviewing trigger queries to refine the skill description, ensuring it activates for the right user intents more often during interactions.

How do I iterate on a skill from benchmark and evaluation feedback?

Iterating from evaluation feedback requires using grader output and benchmark notes to spot weak instructions, flaky tests, or missing edge cases, allowing you to refine the skill until it performs better.

Can I package a validated skill into a distributable file for reuse?

Yes, you can package a validated skill into a distributable skill file, bundling the scripts, references, and assets together for reuse once the evaluation and trigger tuning workflows are complete.

Do I need pyyaml to run skill benchmarking and evaluation workflows?

Yes, pyyaml is required as a dependency to support the skill benchmarking, evaluation, and packaging workflows that utilize scripts and references within the skill creation environment.