// skill.md

Build Claude Code skill files.

Each skill is its own .md file saved to ~/.claude/skills/. Fill the form, get a file named after your skill, repeat for as many skills as you need.

Skill name *

Required — sets the filename

One-line description

When to trigger

Steps

Arguments (optional — how $ARGUMENTS is used)

Example (optional)

Save to ~/.claude/skills/skill.md — Claude Code picks it up on next launch.Each skill is a separate file. Create as many as you need — deploy.md, review.md, test.md — they won't conflict.

Preview updates as you type

// what are claude code skills

Custom slash commands for Claude Code.

Claude Code skills are individual .md files stored in ~/.claude/skills/. Each file is one skill — named after what it does (deploy.md, review.md, test.md). You can have as many as you need — they don't conflict because each has a unique filename.

Instead of typing the same instructions every session, define them once and invoke them with a short command. Skills make Claude behave like a custom CLI tailored to your exact workflow.

~/.claude/skills/deploy.md

---
name: deploy
description: Deploy the current branch to Vercel preview
---

## When to use
When the user asks to deploy, ship, or preview the
current branch to Vercel.

## Steps
1. Run `git status` to confirm no uncommitted changes
2. Run `vercel` to create a preview deployment
3. Share the deployment URL with the user

## Arguments
The branch name to deploy (optional).
Access via $ARGUMENTS.

// use cases

What can you build with skills?

Deployment automation

Define a /deploy skill that runs your deployment pipeline — check git status, run vercel or fly deploy, and share the URL. One command replaces a five-step workflow.

/deploy feat/new-checkout

Code review workflows

A /review skill that runs linting, checks for common mistakes in your codebase, and generates a summary. Consistent review criteria on every PR.

/review src/components/Button.tsx

Database operations

Define /db:migrate, /db:seed, and /db:studio skills that run the right commands for your ORM. Claude stops confusing Prisma and Drizzle commands.

/db:migrate add-user-table

Project-specific test runners

A /test skill that knows your test framework, coverage flags, and which files to watch. No more explaining 'use vitest not jest' in every session.

/test src/lib/converters/

// faq

Frequently asked questions

Where do I put skill files?

Save them to ~/.claude/skills/ in your home directory. Claude Code scans this folder on launch and makes all skills available across every project. You can also put skills in .claude/skills/ inside a specific project for project-scoped commands.

What is $ARGUMENTS?

$ARGUMENTS is a placeholder that gets replaced with whatever the user types after the skill command. For example, if your skill uses $ARGUMENTS and the user types '/deploy feat/my-branch', Claude receives 'feat/my-branch' as the argument.

How many skills can I have?

There's no documented limit. In practice, keeping your skill library under 20 focused skills gives Claude a clear set of commands to choose from. Too many skills with overlapping descriptions can cause Claude to invoke the wrong one.

Can I share skills with my team?

Yes — commit a .claude/skills/ directory to your repo. Any team member who clones the repo gets the same skills automatically. This is the recommended pattern for team-specific workflows like CI checks and deployment commands.

Do skills work in all Claude interfaces?

Skills are a Claude Code feature (the CLI and desktop app). They don't apply to claude.ai chat or the API. For API use cases, the skill content can be used as a system prompt instead.

Can a skill call other tools or run commands?

Yes. Claude Code can execute bash commands, read files, and use any tool it has access to. A skill can instruct Claude to run npm commands, call APIs, read from the filesystem, or chain multiple operations together.