Anthropic said on October 22 it is donating Agent Skills - the open standard that packages specialized knowledge and workflows for AI agents - to the Agentic AI Foundation (AAIF), the Linux Foundation body it co-founded with Block and OpenAI last year. It is the second major standard Anthropic has moved to the foundation, after the Model Context Protocol (MCP).
According to Anthropic's announcement, more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills since the company published it as an open standard in December 2025. Anthropic says 1 million skills have been published to date and nearly 280 million installed. Those figures are the company's own and have not been independently verified.
What a skill is
A skill, under the standard, is a folder containing a SKILL.md file that teaches an agent a specific task or domain, plus any scripts or reference material the task needs. Anthropic says it developed the concept while working on Claude Code: the model handled general knowledge work well but often lacked the procedural knowledge real business tasks require.
The specification at agentskills.io describes skills as a lightweight, open format for extending agent capabilities with specialized knowledge and workflows. Beyond the required SKILL.md - metadata plus instructions - a skill folder can bundle scripts, reference documents and templates. The stated goal is portability: procedural knowledge and company-, team- or user-specific context packaged into version-controlled folders an agent loads on demand, giving it reusable domain expertise, repeatable workflows and cross-product reuse.
To keep that workable at scale, agents load skills through what the specification calls progressive disclosure, in three stages. At startup an agent reads only each skill's name and description - just enough to know when it might be relevant. When a task matches, the full SKILL.md instructions load into context. Only then does the agent follow the instructions, optionally executing bundled code or loading referenced files. The design lets an agent keep many skills on hand with a small context footprint.
The format escaped its origin quickly. Anthropic says that within weeks of the October 2025 launch, developers had published thousands of skills on GitHub and written loaders for Codex, Cursor and OpenCode - prompting the company to publish Agent Skills as an open standard under its own GitHub organization and domain so any agent could use it.
The MCP precedent
The Agentic AI Foundation is a known quantity for this kind of transfer. Anthropic co-founded it with Block and OpenAI last year, and it is already the home of the Model Context Protocol, Anthropic's earlier open standard for connecting models to external tools and data. MCP's trajectory - from a single-vendor protocol to the default wiring between agents and services - is the template Anthropic now explicitly invokes: "As with MCP," the announcement says, the skills project will prioritize community input and transparent decision-making.
Hosting both standards under one Linux Foundation roof also consolidates the emerging agent stack in one governance venue: MCP handles how agents reach tools and data, while Agent Skills defines how procedural expertise is packaged and shared. Rival labs integrating their competitors' formats is easier to justify when the formats answer to a neutral foundation rather than to a competitor.
What changes hands
The donation includes the Agent Skills name and logo, the GitHub organization, and the agentskills.io domain. Anthropic says the AAIF will provide a neutral home for the standard, "ensuring the community has a say in where Agent Skills go next," and that anyone can continue to use the standard or contribute to it without joining the foundation.
Governance will not fully leave its creators. Anthropic says engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals, with David Soria Parra and Mahesh Murag - both Anthropic - as Lead Maintainers.
What comes next
For skill authors, the practical picture today is write once, run across clients: the specification's client showcase lists integrations from JetBrains' Junie and Google's Gemini CLI to smaller open-source runtimes, alongside the major coding agents Anthropic names. Discord and GitHub now serve as the feedback channels for proposed specification changes, per the announcement.
Anthropic says it has proposed a starting roadmap covering versioned releases, a formal proposal process, and security guidance for people who write and install skills - a notable inclusion given that a skill folder can carry executable scripts alongside its instructions. As with MCP, the company says the project will prioritize community input and transparent decision-making.
The move follows a pattern now common among AI labs: publish a format, let competitors adopt it, then transfer stewardship to a neutral foundation before any single vendor's control becomes a reason to refuse it. With skills already loading into rival agents, Anthropic's announcement formalizes a standard the market had effectively already adopted.