Agent Skills defines an open, portable folder format for giving AI agents reusable capabilities. Each skill centers on a required SKILL.md file and can include scripts, references, and assets, letting compatible clients discover a concise skill description before loading detailed instructions and resources only when needed. The specification describes this progressive-disclosure flow as discovery, activation, and execution, and its official showcase currently lists 46 clients, including Claude Code, ChatGPT and Codex, Cursor, GitHub Copilot, Gemini CLI, VS Code, Goose, and OpenCode. Originally developed by Anthropic and released as an open standard, the project accepts community contributions while Anthropic reviews client showcase submissions. The repository's code and specification use Apache-2.0, while its documentation uses CC BY 4.0. Agent Skills is the format itself: Anthropic Agent Skills is an example-skills collection, Awesome Agent Skills is a curated directory, and AGENTS.md is a separate repository-guidance convention. The project has no hosted SaaS tier or tagged release. The format itself does not add telemetry, but data handling and execution permissions depend on the client that loads a skill.
In October 2026 Anthropic donated Agent Skills—its name and logo, the GitHub organization and the agentskills.io domain—to the Agentic AI Foundation under the Linux Foundation, the same foundation that hosts the Model Context Protocol. Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, had integrated the standard by then. Engineers from Anthropic, Google, OpenAI and other companies maintain it, and the proposed roadmap covers versioned releases, a formal proposal process and security guidance for skill authors and users.