Anthropic Agent Skills is the official public reference repo for Claude Skills: reusable folders of instructions, metadata, scripts, templates, and resources that Claude can load when a task matches the skill description. The repo gives teams concrete SKILL.md examples, the Agent Skills spec, starter templates, document-processing skills, a Claude API documentation skill, and example workflows for design, web testing, internal communications, MCP building, and skill creation.
For Claude Code, the repository also acts as a plugin marketplace named anthropic-agent-skills. Teams can add the marketplace from anthropics/skills and install bundled plugin sets such as document-skills, example-skills, or claude-api. That makes it more than a generic prompt collection: it is Anthropic's working distribution surface for turning repeatable guidance into filesystem-backed capabilities that can be installed, triggered on demand, and versioned through a normal repository workflow.
The safest positioning is official Claude skill examples and plugin marketplace, not a neutral third-party prompt marketplace. The repository is public and high-traction, but GitHub still reports no top-level license and Anthropic's README uses mixed licensing language, with many skills Apache-2.0 and document skills described as source-available rather than open source. Anthropic's docs also say Agent Skills are not eligible for Zero Data Retention, so teams should review skill instructions, scripts, and resources before installing untrusted bundles.
