Test-driven development is a way to build software by writing tests first. The same idea works for writing skills for AI agents. You create a test scenario, watch the agent fail, then write the skill to make it pass.
Skills are reference guides that help AI agents follow the right patterns. This approach ensures your skill teaches the correct behavior. You observe the agent without the skill, then add the skill and check again.
Anyone who builds tools for AI agents will benefit. The method is clear and repeatable. It stops you from guessing what works.
Global
mkdir -p ~/.claude/skills/writing-skillsProject
mkdir -p .claude/skills/writing-skillsSource Repository
Setup Matt Pocock Skillsmattpocock/skills
Set up your repo for engineering skills with just a few guided steps
Lark Skill Makerlarksuite/cli
Create reusable skills to automate Feishu tasks with lark-cli commands
Teachmattpocock/skills
Teach yourself any skill with guided lessons and mission-driven practice
Write A Skillmattpocock/skills
Build new agent skills with clear structure and bundled resources
Writing Great Skillsmattpocock/skills
A guide for writing predictable AI agent skills with clear principles and reliable behavior
Ask Mattmattpocock/skills
Pick the perfect skill or flow for your project every time
Writing Skillsobra/superpowers
Test your agent skills like code to make them truly effective
Animejsheygen-com/hyperframes
Safely add Anime.js animations to your HyperFrames compositions with seek control