Anthropic hands Agent Skills to the Linux Foundation's Agentic AI Foundation

On October 22, Anthropic announced that it is donating Agent Skills to the Agentic AI Foundation (AAIF), a fund hosted by the Linux Foundation. Agent Skills is the open format for packaging instructions, scripts and reference files so an AI agent can repeat a task the same way every time. If you have ever dropped a SKILL.md file into a project for Claude Code, this is that format.

What moves to the foundation

Anthropic is handing over the Agent Skills name and logo, the project's GitHub organization, and the agentskills.io domain. AAIF is where Anthropic already placed the Model Context Protocol (MCP), and Anthropic co-founded the foundation with Block and OpenAI. The two standards that shape how most coding agents connect to tools (MCP) and learn procedures (Skills) now sit under the same neutral home.

Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals. The lead maintainers are David Soria Parra and Mahesh Murag. Anthropic says you don't need to join AAIF to use the standard or contribute to it, and feedback on proposed spec changes goes through the project's Discord.

How big is it?

Anthropic first built skills while working on Claude Code. It found that agents were good at general knowledge work but lacked the step-by-step procedural knowledge that real business tasks need. Skills launched in October 2025, and developers quickly built loaders for Codex, Cursor and OpenCode. Anthropic then published Agent Skills as an open standard in December 2025.

Anthropic now reports that more than 70 agent platforms support the format, including ChatGPT, Cursor, GitHub Copilot and VS Code. It also cites roughly one million published skills and nearly 280 million installs. As AI 360 points out, those figures come from Anthropic and haven't been independently audited.

What comes next

The announcement proposes three early priorities but gives no dates:

• Versioned releases of the specification
• A formal proposal process for changes
• Security guidance for people who write and install skills

The last one matters. A skill can bundle executable scripts alongside its instructions. That means a skill pulled from the internet is effectively code your agent may run, and AI 360 flags the open questions around provenance, permissions, versioning and vulnerability disclosure.

The move was not a surprise. Simon Willison noted that when Anthropic first called skills an open standard last December, he wrote that he expected them to end up at AAIF.

Why it matters if you're learning to build with AI

For learners, the practical upshot is portability. A skill you write today in .claude/skills/ follows a spec that dozens of other tools also read. Neutral governance makes it more likely that the format stays stable and that you won't have to rewrite your workflows when you switch agents. Two habits are worth building now. Stick to the core spec fields when you want a skill to travel between tools. And treat third-party skills, especially ones that ship scripts, with the same caution you'd give any dependency until the promised security guidance lands.