Agent Skills
Claude Code Agent Skills
How skills load in Claude Code, how to install Anthropic's published set, and how to write your own SKILL.md, checked against Anthropic's skills repository and the Claude Code documentation.
Updated
Claude Code skills are folders of instructions, scripts and resources that Claude loads when a task needs them. Each skill is a SKILL.md file with YAML frontmatter, a name and a description, followed by the instructions. You install Anthropic's published skills through the plugin marketplace, or write your own under ~/.claude/skills or a project's .claude/skills directory.
Install Anthropic's skills in Claude Code
Anthropic publishes its example skills as a Claude Code plugin marketplace called anthropic-agent-skills. Register the repository once, then install a plugin: document-skills bundles the Excel, Word, PowerPoint and PDF skills, and example-skills bundles the creative, development and communication skills. After installing, you use a skill by mentioning it in a request.
Register the marketplace
/plugin marketplace add anthropics/skillsInstall a plugin
/plugin install document-skills@anthropic-agent-skills
/plugin install example-skills@anthropic-agent-skillsCommands from the anthropics/skills README, read 2026-09-23. The same repository also publishes claude-api and academy-guide plugins.
Skills in Anthropic's repository
Anthropic's repository holds 19 skills. The four document skills (docx, pdf, pptx and xlsx) power Claude's own file features and are source-available rather than open source; most of the rest are open source under Apache 2.0. The cards below cover the ones most people install first, with each skill's description paraphrased from its SKILL.md.
PDF
Reads and extracts text and tables from PDFs, merges and splits files, rotates pages, adds watermarks, fills forms and creates new PDFs.
Word (docx)
Creates, reads and edits Word documents and templates, including formatting such as tables of contents, headings and page numbers.
Excel (xlsx)
Opens, edits, cleans and creates spreadsheets, including formulas, formatting and charts, for .xlsx, .xlsm, .csv and .tsv files.
PowerPoint (pptx)
Builds, reads and edits slide decks and templates whenever a .pptx or .potx file is the input or the output.
Frontend Design
Guidance for distinctive, intentional visual design when building or reshaping a UI: aesthetic direction, typography and avoiding templated defaults.
MCP Builder
A guide to building MCP servers in Python with FastMCP or in TypeScript with the MCP SDK, from tool design to evaluation.
Skill Creator
Creates and edits skills, runs evals against them, benchmarks performance and tunes a skill's description so it triggers reliably.
Webapp Testing
Tests local web applications with Playwright: checks frontend behaviour, debugs the UI, captures screenshots and reads browser logs.
Web Artifacts Builder
Builds multi-component claude.ai HTML artifacts with React, Tailwind CSS and shadcn/ui when an artifact needs state or routing.
Create your own Claude Code skill
A skill is a directory containing a SKILL.md file. The YAML frontmatter tells Claude when to use it, and the markdown body holds the instructions Claude follows when it runs. The directory name becomes the command you type, and the description lets Claude load the skill automatically when a request matches. The body only loads when the skill is used.
- 01
Create the skill directory
Personal skills live under ~/.claude/skills and are available in every project on your machine. Use a project's .claude/skills directory instead to share the skill with everyone who clones the repository.
mkdir -p ~/.claude/skills/summarize-changes - 02
Write SKILL.md with a name and a description
The frontmatter needs a name, lowercase with hyphens, and a description that says what the skill does and when to use it. Put the key use case first in the description, because that text is what Claude reads when deciding whether to load the skill.
--- name: summarize-changes description: Summarizes uncommitted changes and flags anything risky. Use when the user asks what changed or wants a commit message. --- # Summarize changes [Instructions Claude follows when this skill runs] - 03
Invoke it by name, or let Claude pick it
Type /summarize-changes to run the skill directly, or ask something that matches its description and Claude loads it on its own. Add disable-model-invocation: true to the frontmatter when a skill should only ever run when you call it.
/summarize-changes
Where Claude Code loads skills from
Where you save a skill decides which sessions load it. Personal skills follow you across projects on one machine, project skills travel with the repository, plugin skills arrive with an installed plugin under a plugin:skill namespace, and enterprise skills are deployed by an organisation through managed settings. When two skills share a name, the location decides which one runs.
| Location | Path | Loads in |
|---|---|---|
| Personal | ~/.claude/skills/<skill-name>/SKILL.md | All your projects on this machine |
| Project | .claude/skills/<skill-name>/SKILL.md | Sessions in this repository; commit it to share it |
| Nested | <subdir>/.claude/skills/<skill-name>/SKILL.md | Sessions started in or below that directory |
| Plugin | <plugin>/skills/<skill-name>/SKILL.md | Wherever the plugin is enabled, as /plugin-name:skill-name |
| Enterprise | .claude/skills/<skill-name>/SKILL.md in the managed settings directory | All users on machines where the organisation deploys it |
Locations from the Claude Code skills documentation, read 2026-09-23.
Related guides for Claude Code
Skills teach Claude Code how to do a task; the pages below cover the pieces around them. They explain how to install SpringBrand in the Claude Code CLI, how Claude connects to MCP servers, how the SpringBrand command-line tools work, and how to give an agent access to your GitHub repositories through an official server.
- Install SpringBrand for Claude Code CLI
Step-by-step instructions to install and connect SpringBrand with Claude Code CLI.
- Claude MCP
How Claude MCP works on three surfaces: claude mcp add in Claude Code, .mcpb extensions in Claude Desktop, and remote custom connectors on claude.ai and Cowork.
- SpringBrand CLI Quickstart
Use SpringBrand from your terminal: install the plugin in Claude Code CLI or Codex CLI, sign in with one command, and verify the connection before running tasks.
- GitHub MCP Server
How to install GitHub MCP Server: the remote endpoint at api.githubcopilot.com, the local Docker image, PAT and OAuth auth, and the toolsets to turn on.
Claude Code skills questions
These answers cover what people ask when they start using skills in Claude Code: how a skill differs from CLAUDE.md and custom commands, whether skills work outside Claude Code, which fields SKILL.md needs, how to stop Claude loading a skill automatically, and whether Anthropic's own skills are free to use.
How is a skill different from CLAUDE.md?
CLAUDE.md content is loaded into every session. A skill's body loads only when the skill is used, so long reference material or a multi-step procedure costs almost nothing until you need it. Anthropic's docs suggest turning a section of CLAUDE.md into a skill once it has grown into a procedure rather than a fact.
What happened to custom slash commands?
They were merged into skills. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy and work the same way, and existing command files keep working. Skills add a directory for supporting files, frontmatter that controls who can invoke them, and automatic loading.
Do skills only work in Claude Code?
No. Claude Code skills follow the Agent Skills open standard published at agentskills.io, which other AI tools also support. Anthropic's example skills are also available on paid Claude.ai plans, and the Claude API accepts Anthropic's pre-built skills as well as custom ones you upload.
Which fields does SKILL.md need?
Anthropic's repository says the frontmatter needs only a name, a unique lowercase identifier with hyphens, and a description of what the skill does and when to use it. Claude Code adds optional fields, such as disable-model-invocation and allowed-tools, on top of the standard.
Are Anthropic's skills open source?
Most are, under Apache 2.0. The four document skills, docx, pdf, pptx and xlsx, are source-available rather than open source: Anthropic shares them as a reference for complex skills used in production. Anthropic also notes the skills are for demonstration, so test them before relying on them.
Skills teach an agent a task
A skill gives Claude Code the instructions for a task; it does not give it access to the systems where the task has to happen. SpringBrand connects agents to real-world tools, so a skill's instructions can end in finished work: a published page, an updated record or a delivered report.