Guide · GitHub data checked November 11, 2026
What skills are, how to install them in Claude Code, Claude.ai and the API, how to write your own, and the open-source skill libraries worth knowing, grouped by what you want Claude to do.
In one sentence
A Claude skill is a folder with a SKILL.md file that teaches Claude a specific job. Claude keeps the short description in mind and loads the full instructions only when a task needs them.
---
name: brand-voice
description: Rewrite drafts in our brand voice.
Use when the user asks to edit, polish or
write customer-facing copy.
---
# Brand voice
1. Short sentences. No jargon.
2. Lead with the customer's outcome.
3. Check every claim against facts.md.
## Files
- facts.md: approved product facts
- examples.md: before and after editsEach skill lives in its own folder. The only required file is SKILL.md, which starts with two fields in YAML frontmatter: name, a unique identifier in lowercase with hyphens, and description, what the skill does and when to use it. The instructions follow in plain Markdown.
A folder can also hold scripts Claude can run and reference files it can read. Because Claude only loads the full content when a task matches the description, you can install many skills without filling the conversation with instructions you are not using.
Claude Code
For every project: ~/.claude/skills/skill-name/SKILL.md. For one project, shared with your team in git: .claude/skills/skill-name/SKILL.md.
Or install a plugin that bundles skills, for example Anthropic's:
/plugin marketplace add anthropics/skills
/plugin install document-skills@anthropic-agent-skillsClaude.ai
On paid plans, Anthropic's example skills are available in Claude.ai, and you can upload your own skill as a folder. Anthropic's help article "Using skills in Claude" has the current steps for your plan.
Claude API
Developers can use Anthropic's pre-built skills and upload custom skills through the Claude API. The Skills API quickstart in Anthropic's documentation shows the request format.
The fastest start is Anthropic's skill-creator skill, which interviews you and drafts the folder. Whether you use it or write by hand, these rules make the difference between a skill that fires and one that sits unused.
They solve different problems and work well together: MCP gives Claude access to your systems, skills tell it how to do the job, subagents split the work.
| Feature | What it is | When it runs | Best for |
|---|---|---|---|
| Skill | A folder with a SKILL.md file, and optionally scripts and reference files | Claude reads the short description at all times and loads the full instructions only when the task matches | Repeatable know-how: a house style, a review checklist, a document format |
| MCP server | A connector that follows the Model Context Protocol | Claude calls its tools when it needs data or an action in another system | Access: reading a CRM, posting to Slack, querying a database |
| Slash command | A saved prompt in Claude Code | You trigger it yourself by typing /name | A shortcut you run on demand |
| Subagent | A separate Claude Code agent with its own instructions, context and tools | Claude delegates a task to it and gets the result back | Isolating a long or specialised task from the main conversation |
| CLAUDE.md | A project instructions file | Loaded at the start of every session in that project | Rules that always apply: conventions, commands, things to avoid |
Open-source skill libraries selected for usefulness and activity. Stars and last update come from the GitHub API on November 11, 2026. We did not write these projects: each one belongs to its authors.
anthropics
Anthropic's own example skills, including the document skills (docx, pdf, pptx, xlsx) that power file creation in Claude, plus skill-creator, frontend-design and mcp-builder.
anthropics
The Anthropic-managed directory of Claude Code plugins, which can bundle skills, commands and agents.
obra
A skills framework and development method for coding agents: brainstorming, planning, test-driven development and code review as reusable skills.
affaan-m
An agent harness with skills, memory, security checks and research-first development habits for Claude Code and other coding agents.
wshobson
Plugins, agents, skills and commands for Claude Code and other coding CLIs, written once and shared across tools.
nextlevelbuilder
A design-intelligence skill for building professional UI and UX across platforms: styles, palettes, font pairings and layout rules.
Nutlope
A design skill that steers Claude Code, Cursor and Codex away from generic, AI-looking interfaces.
coreyhaines31
Marketing skills for Claude Code and other agents covering conversion optimisation, copywriting, SEO, analytics and growth.
blader
A skill that edits text to remove common signs of AI-generated writing.
mvanhorn
Researches a topic across Reddit, X, YouTube, Hacker News and the web, then writes a grounded summary of the last 30 days.
K-Dense-AI
A large library of skills for scientific work: literature, data analysis and domain tools for research teams.
Imbad0202
Skills that walk Claude Code through academic research: research, write, review, revise and finalise.
tt-a1i
Turns an idea, plan or codebase into an interactive diagram, as a skill for Claude Code and other agents.
Graphify-Labs
Turns a codebase with its docs, schemas and configs into a knowledge graph you can query from your agent.
obsidianmd
Official Obsidian skills that teach an agent to work with Obsidian files, Bases and JSON Canvas.
JuliusBrussee
A skill and proxy that makes coding agents answer in far fewer words, to save tokens.
mukul975
Structured cybersecurity skills mapped to frameworks such as MITRE ATT&CK and NIST CSF, for defensive work.
alirezarezvani
A large MIT-licensed library of skills, agents and commands across engineering, marketing, product, compliance and C-level advisory.
sickn33
A very large collection of reusable agent skills for Claude Code, Codex, Cursor and Gemini CLI, with local catalog discovery.
ComposioHQ
A curated list of Claude skills, resources and tools.
VoltAgent
A curated collection of agent skills from official dev teams and the community, compatible with several coding agents.
travisvn
A curated list of Claude skills and resources, with a focus on Claude Code.
davila7
An open-source registry and CLI to install Claude Code agents, commands, skills, hooks and MCP servers.
A skill can include scripts that Claude runs on your machine or in your environment. Treat it like any code you download.
We build automations on n8n where a Claude step follows a skill, so every draft, summary or classification follows the same rules, with a person approving anything sensitive. The workflows run in your own accounts.
Claude skills are folders of instructions, scripts and resources that Claude loads when they are relevant to a task. Each skill has a SKILL.md file with a name and a description in its frontmatter, followed by instructions. They teach Claude to do a specific job the same way every time, such as producing documents in your brand style or following your review checklist.
In Claude Code, put a skill folder in ~/.claude/skills for all your projects or in .claude/skills inside a project, or install a plugin that bundles skills. Then simply ask for the task: Claude picks the skill whose description matches. On paid Claude.ai plans you can turn on Anthropic's skills and upload your own, and developers can use skills through the Claude API.
A skill is know-how that Claude loads into its current work. A subagent is a separate agent with its own context and tools that Claude hands a task to. Skills make Claude better at a task; subagents split work across agents. They are often used together.
A slash command runs only when you type it. A skill is picked up by Claude when the task matches its description, so you do not have to remember to call it.
skill-creator is a skill in Anthropic's public skills repository that helps you write new skills: it asks what the skill should do and drafts the SKILL.md and folder structure with you.
The skills themselves are plain files. Anthropic's repository and most community libraries on GitHub are open source, under licences such as MIT or Apache 2.0. You still need a Claude plan or API access to run them.
Treat a skill like any code you download. Skills can include scripts that Claude runs, so read the SKILL.md and the scripts first, prefer sources you trust, and keep secrets out of skill folders.