Short answer
Claude skills are folders that teach Claude how to do a specific, repeatable task. Each skill has a SKILL.md file with a name, a description, and instructions, and it can include scripts, templates, or reference files. Claude only sees the short descriptions until a request matches one, then loads that skill in full. Skills work in the Claude app, Claude Code, and the Claude API, and follow the open Agent Skills standard, so the same folder can also work in tools such as Cursor and Codex.
Words used in this guide
- Workflow: a set of steps used to finish a task.
- Repository or repo: a project folder whose changes are saved and tracked.
- Git: a tool that records file changes so people can review or undo them.
- MCP: a standard way for an AI tool to connect to other tools and information.
What this means
How a Claude skill works
A skill is a folder such as brand-guidelines/ containing SKILL.md. The top of SKILL.md is YAML frontmatter with a name and a description; the rest is Markdown instructions. At the start of a conversation Claude sees each enabled skill's name and description. When your request matches a description, Claude reads the full SKILL.md and any files it points to. Because the full text loads only when needed, you can have many skills without filling the context window. For every frontmatter field and limit, see the SKILL.md frontmatter reference.
What a skill can contain
Instructions and checklists for a task, example inputs and outputs, templates (for example a slide or document template), reference material such as a style guide or API notes, and scripts that Claude can run with code execution. Keep SKILL.md focused and move long reference material into separate files that SKILL.md links to.
Built-in skills from Anthropic
Anthropic provides skills for creating and editing Excel spreadsheets, Word documents, PowerPoint presentations, and PDFs. With code execution and file creation turned on, Claude uses them automatically when your request needs them. Anthropic also publishes example skills (creative, technical, and enterprise workflows) and a skill-creator skill in the anthropics/skills GitHub repository.
Examples of skills people build
Marketing: apply brand voice and formatting rules to every draft. Operations: turn meeting notes into a standard status report. Documents: fill a company template or check a contract against a checklist. Engineering: review a pull request, write release notes, or follow a deployment runbook. Data: clean a CSV and produce the same chart every week. Good skills do one job and say clearly when they should be used.
Skills vs prompts, projects, CLAUDE.md, and MCP
A prompt is text you paste each time; a skill is saved and loaded automatically when relevant. CLAUDE.md (in Claude Code) is always loaded project context, while a skill loads only for its task. MCP connects Claude to outside tools and data, while a skill teaches Claude how to do a task, and a skill can tell Claude to use an MCP tool. Use a skill whenever you keep repeating the same instructions. In Claude Code, see skills vs MCP vs subagents vs hooks for when to use each.
Where Claude skills work
In the Claude app (claude.ai and Claude Desktop) on Free, Pro, Max, Team, and Enterprise plans, with code execution enabled; you manage them in Customize > Skills. In Claude Code, skills live in ~/.claude/skills/ or a project's .claude/skills/, and can also arrive through plugins. In the Claude API, skills are available with the code execution tool. Skills you enable in the Claude app also sync into Claude Code when you sign in with the same account. For the exact folders on disk, see where are Claude skills stored?.
Only install skills you trust
A skill can include scripts and instructions that Claude follows, so a malicious skill could try to run unwanted actions or leak data. Anthropic's guidance is to install skills only from trusted sources and to read a skill's files, especially scripts and any network access, before enabling it.
Simple steps
- Try Anthropic's built-in skills: turn on code execution and ask Claude to create a spreadsheet or slide deck.
- Pick one task you repeat every week.
- Find an existing skill for it, or write a folder with a SKILL.md that has a clear name and description.
- Install it: upload a ZIP in Customize > Skills (Claude app) or copy it to ~/.claude/skills/ (Claude Code).
- Ask Claude for that task and check the result on a small example.
- Improve the description if Claude does not pick the skill up, then share it with your team.
Where to go next
If you want to try this in a real project, start with how to use AIPM and the command reference. If you want to share your own workflow, read the publishing guide.
FAQ
Are Claude skills free?
Skills are available on the Free, Pro, Max, Team, and Enterprise plans. They need code execution to be enabled. Many community skills are open source; check each skill's license.
What is the difference between Claude skills and MCP?
MCP gives Claude a connection to an outside tool or data source. A skill gives Claude instructions and files for a task. They work together: a skill can describe how and when to use an MCP tool.
Do I need to code to use Claude skills?
No. You can enable Anthropic's skills with a toggle and upload a skill as a ZIP in the Claude app. Writing a skill is mostly writing clear instructions in Markdown; scripts are optional.
What is the skill-creator skill?
skill-creator is an Anthropic skill that helps you write and test new skills. In Claude Code you can install it with /plugin install skill-creator@claude-plugins-official.
Are Claude skills the same as agent skills?
Claude skills follow the open Agent Skills standard (agentskills.io). Claude Code adds some extra frontmatter fields, but a skill that uses the standard fields can also work in other agents that support the standard.