// Skill profile
anthropics/skills is Anthropic's official public repository of Agent Skills for Claude. A skill is a folder containing a SKILL.md instruction file plus optional scripts and resources that Claude loads dynamically to perform specialized tasks. The repo ships example skills across four categories — Creative and Design (art, music, design), Development and Technical (testing web apps, MCP server generation), Enterprise and Communication (communications, branding), and Document Skills (docx, pdf, pptx, xlsx) — alongside the Agent Skills specification (./spec) and a starter template (./template) for building custom skills. Skills install into Claude Code via the plugin marketplace (`/plugin marketplace add anthropics/skills`), and are also available on Claude.ai (paid plans) and through the Claude API Skills endpoint. The document skills (docx, pdf, pptx, xlsx) are source-available as production references; the rest are Apache 2.0. With 186K stars, this is the most-starred skills repository on GitHub and the canonical reference for the SKILL.md format that the broader ecosystem (obra/superpowers, mattpocock/skills, and 100+ community skill packs) builds on.
anthropics/skills is the official Anthropic-maintained collection of Agent Skills for Claude. Where community skill packs build on top of the SKILL.md format, this repository defines it — shipping the Agent Skills specification alongside production-quality example skills that demonstrate the pattern across creative, technical, enterprise, and document-processing use cases. With 186K stars it is the most-starred skills repository on GitHub, reflecting its role as both a usable skill library and the canonical reference implementation for the broader skills ecosystem.
The repository serves three audiences: developers who want ready-made skills for Claude, skill authors who need the specification and starter template, and teams evaluating agent skill architectures who want to see how Anthropic structures production skills internally.
These four document skills power Claude’s built-in file capabilities on Claude.ai and are shared as source-available references for teams building complex production skills.
Every skill follows a consistent format — a folder containing a SKILL.md with frontmatter:
---
name: my-skill-name
description: A clear description of what this skill does and when to use it
---
# My Skill Name
[Instructions Claude follows when this skill is active]
## Examples
- Example usage 1
## Guidelines
- Guideline 1
The name field is a unique lowercase identifier (hyphens for spaces). The description tells Claude when to activate the skill based on context.
Claude Code — register the repo as a plugin marketplace, then install skill packs:
# Add the Anthropic skills marketplace
/plugin marketplace add anthropics/skills
# Install document skills
/plugin install document-skills@anthropic-agent-skills
# Install example skills
/plugin install example-skills@anthropic-agent-skills
Claude.ai — Example skills are available to paid plan users. Custom skills can be uploaded following the official help article.
Claude API — Pre-built skills are available through the Skills API endpoint. Custom skills can be uploaded programmatically.
This repository is the upstream source of truth for the SKILL.md format. Every major community skill pack — obra/superpowers (276K stars), mattpocock/skills (20K stars), and hundreds of others — builds on the patterns defined here. Understanding how Anthropic structures skills (frontmatter schema, instruction format, example patterns, test conventions) is essential context for anyone building or evaluating agent skills.
The spec directory (./spec) documents the format formally, while the template directory (./template) provides a ready-to-use scaffold for new skills.
THIRD_PARTY_NOTICES.mdAI agents that work well with Anthropic Skills.
Single 70-line CLAUDE.md derived from Karpathy's LLM coding-pitfall observations — #1 GitHub trending for 28 consecutive days.
A complete software development methodology and composable skills framework for coding agents — brainstorm, plan, execute, review, ship.
Turn any codebase into a queryable knowledge graph — deterministic AST parsing via tree-sitter, Leiden clustering, no vector store required.