What Are Claude Skills? One File That Runs a Whole Method
TL;DR
A skill is a folder named after the skill, with one file inside it called SKILL.md. The top says what it does and when to use it. The rest is the method.
A skill beats a plain prompt because it carries a tested process. Same input, same output, no matter who runs it or when.
Run skills two ways: toggle and upload in the Claude desktop app, or drop a folder on disk in Claude Code. Build with skill-creator, vet anything from outside before it runs.
Most people use Claude by typing a prompt and hoping. A skill is the fix.
A Claude skill is a folder with one file inside it, called SKILL.md. The top of the file says what it does and when to use it. The rest is the method. You say the trigger, the skill fires, and the model runs a tested process instead of winging it.
That is the whole idea. Let me show you why it matters and how to set one up.
A skill is packaged know-how
Think of a skill as a recipe card the model picks up the moment your request matches what is written on the front.
It sits quiet. You ask for something. The right skill recognises the request, fires, and runs a full method behind it. Same result every time.
The file is plain markdown. A name, a description, then the steps. No app to install. No code to wire up. A folder named after the skill, a SKILL.md inside it, and you have a working skill.
Why a skill beats just asking
Ask Claude a plain question and it improvises a fresh answer every time. Useful for one-offs. A problem for anything you repeat, because the output drifts.
A skill carries a tested process. So the output is the same no matter who runs it or when. That is the gap between "the AI had a go" and "the AI ran our process".
And skills stay cheap. The model first sees only the name and the one-line description. It reads the full method just when the skill fires, and opens any extra files only if the method points at them. So a skill can hold a lot of detail without slowing down every chat.
The one file: SKILL.md
The most important line in the whole file is the description. It is what the model reads to decide when to fire the skill. Say what it does and when to use it. A vague description, and the skill never fires at the right time.
That is the floor: a folder, a SKILL.md, a sharp description, and the method below it. When the method gets long, you split it. Add a references folder for docs, a scripts folder for code, an assets folder for templates. The skill reads them only when it needs them, so the main file stays short.
Two ways to run them
There are two places you set skills up in Claude, and they keep separate stores.
The desktop app and claude.ai. Manage skills in a settings panel. Toggle Anthropic's built-in ones on, or upload your own as a zip. Best for everyday work where you do not want to touch the file system.
Claude Code. A skill is just a folder on disk: drop it in .claude/skills/ and Claude finds it. Version it with your code, share it with the team. Best for builds and serious work. The same SKILL.md format is also read by other tools, so a skill you write is not locked to one place.
Build one, or find a proven one
Two skills do the heavy lifting. skill-creator is Anthropic's own skill for building skills. It walks you from idea to a finished SKILL.md, runs evals, and tunes the description so the skill fires at the right time.
Before you write one from scratch, check if a proven one exists. The open directory at skills.sh ranks skills by installs, and Anthropic's public skills repo holds the standard plus examples. One rule though: a downloaded skill is code that runs inside your agents, so vet anything from outside before it touches a live setup.
Where skills fit in the bigger picture
Skills are the muscle of an AI Operating System. On their own they package your methods. Wired into a memory that holds your context, they turn what you know into finished work. That is the difference between using AI and running it, which is the whole idea behind the AI Operating System.
And once you run a few skills across more than one project, you need to track them, or you forget what you have and build the same thing twice. That is what building your own AI stack is for. If you are weighing whether to build an agent to run those skills, DevWiz's guide to building an AI agent is a good next read.
See the full guide
This is the short version. The full walkthrough covers what a skill is, the five-plus tools that read SKILL.md, installing in the desktop app and Claude Code, the right folder structure, the front matter rules, how to vet a skill, and where to find good ones.
Read the Agent Skills guide for the complete build. Then, if you want to know which part of an AI Operating System will move the needle first for your business, run the assessment. Five minutes, scored, with the first move for your situation.
Frequently Asked Questions
James Killick
Founder
The AI Orchestrator. 10+ years building digital products and 200+ apps shipped, now helping $1M+ educators and consultants turn their IP into AI-powered delivery systems.
James Killick founded and runs The AI Orchestrators.
More from James Killick