The hands-on way to learn Claude Code skills and Agent Skills: guided paths, real lessons, and a working skill you build as you go.
Short, plain-language lessons: what a skill is and why it works the way it does.
Author a working SKILL.md as you go. Every path produces a tangible artifact.
Make it trigger reliably, package it, and hand it off: the full lifecycle.
Start from zero. Understand what a Claude skill is, read a real SKILL.md, then write, invoke, and verify your own, with no prior skill experience needed.
You'll be able to: You can author a minimal skill and confirm Claude invokes it on the right request.
The difference between a skill that helps and one that gathers dust is triggering. Go deep on descriptions, frontmatter, scoping, and progressive disclosure so your skills fire exactly when they should.
You'll be able to: You can write descriptions and structure files so a skill loads on the right task and stays out of the way otherwise.
Turn a working skill into something you can bundle scripts with, run across tools, and hand to a teammate or the community, with versioning that does not break their setups.
You'll be able to: You can ship a skill that bundles executable helpers and distribute it without surprising its users.
Each path ends with something you , and a skill you can keep using.
The paths end with a working skill, and working skills deserve an audience. Submit yours to the community: we read every one, feature the strongest on the blog, and pass catalog-worthy skills to our sibling directory. Video walkthroughs of real skill builds are on the way too.
New lessons and paths, in your inbox. No spam. We email when a new path or lesson ships, nothing else.