What Are Claude Skills? A Plain English Guide

Learning how to use Claude skills comes down to two things: triggering the ready-made ones when you need them, and building your own so Claude does your repetitive work your way. A skill is a reusable set of instructions Claude follows to complete a specific task, and once you've set one up you run it with a single command instead of re-explaining the job every time. This guide walks through both, step by step, with no coding required.

We've taught this to thousands of people, and the same small mistakes come up every time. The steps below are the order we teach it in, including the one habit that makes the difference between a skill that works and one that frustrates you.

Watch: the full walkthrough is in our free Claude Cowork course, which covers skills from first use to building your own.

Step 1: Turn on the Skill Creator

Before anything else, switch on the one skill everyone should have: Anthropic's Skill Creator. Find it under Customize, then Skills, then browse the available skills, and enable it. It's the skill you use to build every other skill, and it should be one of the first things you turn on.

While you're there, you'll see the pre-built skills Anthropic ships, including ones for creating Word, Excel, PowerPoint, and PDF files. Those work automatically whenever they're relevant, so you don't need to do anything to "use" them beyond asking for the output.

Step 2: Trigger a skill

Using an existing skill is the easy part. In a new task, type / and pick the skill from the list, or simply describe what you want in a way that matches the skill's purpose. Skills carry a short description telling Claude when they apply, so a well-written one triggers on the right request without you naming it.

For example, an email-triage skill is written to fire when you ask things like "check my inbox" or "what needs a reply today." You don't have to remember the skill's exact name; you just ask for the outcome, and Claude reaches for the matching skill.

A quick tip: keep an eye on which skill Claude picks, especially early on. If two skills have similar descriptions, the clearer one wins, which is a good reason to name and describe your own skills precisely.

Step 3: Build your own skill the right way

This is where most people go wrong. They open the Skill Creator and say "build me a skill that does X." Don't. The reliable way is to do the task with Claude first, then save it.

Walk Claude through the job step by step, because you're the expert and you're teaching it your method. Use real inputs: a real draft, a real spreadsheet, a real folder. Check the result and refine it until it's right. Only once the output is correct do you turn it into a skill, using the Skill Creator or the "turn this into a skill" option. The Skill Creator even tests what it builds before saving, so a skill that resizes an image, say, gets checked against both a short and a long input.

The reason this works is simple. You're not asking Claude to imagine a process; you're capturing one that already produced the right result. The skill becomes a recording of what worked, not a guess.

Step 4: Know what's inside a skill

You don't need to write code, but it helps to know what the Skill Creator packages up, because it tells you what to check. A skill is a folder with three kinds of content.

The instructions live in a file called SKILL.md, written in plain markdown. Supporting resources sit alongside it: templates, fonts, logos, or saved scripts the skill needs. More advanced skills can also bundle their own helper agents, the way our video-clipping skill includes a separate agent that analyses footage frame by frame.

When you review a freshly built skill, open it and check the pieces are there. We once built a skill that quietly left out a key image file, which only showed up when someone else tried to run it. A quick look at the contents catches that.

Step 5: Test it in a fresh task

Always test a new skill in a clean task, not the one you built it in, because the building conversation already has all the context loaded. Hand it a realistic input and check the output end to end.

When it works, you're done: the job that took a careful back-and-forth the first time now runs from one command. Feed a skill a list of fifteen items and it produces fifteen correct outputs without you touching the process again.

Step 6: Update and share skills

Skills aren't fixed. To change one, use the same method you used to build it: do the new version of the task with Claude first, confirm it works, then update the skill and save over the old one. To add an option, for example a dark-mode version of an image, get Claude to produce it once, verify it, then bake the choice into the skill so it asks which one you want next time.

Sharing depends on your plan. On a Team or Enterprise plan you can share a skill with your whole organisation, and your future updates reach everyone automatically. As an individual, you can download a skill as a file and send it on, though a downloaded copy won't receive your later updates. Either way, this is how one person's expertise becomes something the whole team can run.

Start using skills today

The fastest way to learn how to use Claude skills is to pick one repetitive task you already do, run it with Claude once, and save it. That single loop, do it together, verify, save, is the whole method. From there, every job you explain more than once becomes a skill you never explain again.

If you want to learn this properly and build a skill library around how your team actually works, our hands-on Claude Cowork training takes you from your first task to a set of skills your whole team can run.