Anthropic Skills

Anthropic Skills is Anthropic's official GitHub repository of modular, filesystem-based skill packages that extend Claude's behavior for specialized tasks. Each skill is a folder containing a SKILL.md file with YAML frontmatter (name and description) and markdown instructions that Claude loads dynamically. The repo covers document processing, creative tools, development utilities, and enterprise workflows, and also serves as the reference implementation of the open Agent Skills standard.

In Claude Code, run `/plugin marketplace add anthropics/skills` to register the repo, then browse and install either `document-skills` or `example-skills` from the `anthropic-agent-skills` marketplace entry. On Claude.ai, the example skills are already available to paid plan users with no setup needed. Via the Claude API, use the Skills API (see the Skills API Quickstart in the docs) to upload and use pre-built or custom skills. Once installed, you activate a skill simply by mentioning the relevant task in your message.

Most skills in the repository are open source under the Apache 2.0 license. The exception is the document creation and editing skills (docx, pdf, pptx, xlsx), which are source-available but not open source. Anthropic included them as a reference for production-grade skill design. The repository itself is free to clone and use.

Skills are best suited for repeatable, domain-specific workflows you run across many Claude conversations: enforcing brand guidelines, processing documents at scale, automating multi-step development tasks, or packaging organizational knowledge so teams share a consistent Claude behavior. The platform docs describe skills as the alternative to prompts when you need load-on-demand expertise rather than one-off instructions.

Skills and system prompts solve different problems. System prompts are conversation-level, one-off instructions that must be repeated every session. Skills are filesystem-based resources Claude loads on-demand using a three-level progressive disclosure model (metadata always, instructions when selected, scripts only when needed), which keeps context consumption low and eliminates repetition. Per the platform docs, skills are designed to compose, letting you combine multiple skills to build complex workflows.

The README is explicit: skills in this repository are provided for demonstration and educational purposes only, and behavior you get from Claude may differ from what the skill definitions show. Agent Skills are also not eligible for Zero Data Retention, meaning data is retained under Anthropic's standard retention policy. Always test custom skills thoroughly in your own environment before relying on them for critical or sensitive tasks.