Agent Skills Moves to the Agentic AI Foundation: What the Donation Changes for Builders
Anthropic donated the Agent Skills open standard to the Linux Foundation's Agentic AI Foundation on October 22, alongside Open Responses, OpenEnv and OpenSandbox. What the announcements say, what is still a roadmap, and what to check before you depend on it.
Anthropic announced on October 22 that it is donating the Agent Skills open standard to the Agentic AI Foundation (AAIF), which sits under the Linux Foundation. The standard stays free to use, and you do not need to join AAIF to use it or contribute. What changes is who governs it. I have not built against the spec for this piece, so everything below comes from the announcements.
I read two primary sources: Anthropic's post, Donating Agent Skills to the Agentic AI Foundation, and the Linux Foundation press release, Agentic AI Foundation Announces Four New Projects. I did not read the specification at agentskills.io.
What a skill is
Per Anthropic, a skill is a folder containing a SKILL.md file that teaches an agent a specific task or domain, plus any scripts or reference material it needs. Anthropic says it developed skills while working on Claude Code, which handled general knowledge work well but often lacked the procedural knowledge that real business tasks need. The Linux Foundation describes the format as one agents can load on demand, so know-how is portable across products.
What Anthropic says happened
These numbers are Anthropic's own:
- More than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills since the standard was published in December 2025.
- Skills launched in October 2025. Within weeks developers published thousands of skills on GitHub and wrote loaders for Codex, Cursor and OpenCode.
- So far 1 million skills have been published and nearly 280 million installed.
I have no independent count for any of these. "Installed" is not defined in the post.
What is being donated
Anthropic says it is donating the Agent Skills name and logo, the GitHub organization and the agentskills.io domain. Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals. David Soria Parra and Mahesh Murag are the Lead Maintainers.
This follows the same path as the Model Context Protocol, which Anthropic donated to AAIF earlier. AAIF was co-founded by Anthropic, Block and OpenAI, per Anthropic.
The roadmap is a proposal
Anthropic says it has proposed starting with a roadmap covering versioned releases, a formal proposal process, and security guidance for people who write and install skills. It says it will take feedback through the project's Discord and GitHub. This is a proposal, not a published schedule. I found no dates in either source.
The security guidance item matters because a skill can include scripts. Neither source says what the guidance will contain.
The other three projects
The press release says AAIF announced four projects the same day, which brings it to 10 hosted projects from three at launch in December 2025:
| Project | Contributed by | What it is, per the release |
|---|---|---|
| Agent Skills | Anthropic | Open format for giving agents new capabilities and domain expertise |
| Open Responses | OpenAI | Open specification for multi-provider LLM interfaces, based on the OpenAI Responses API |
| OpenEnv | Hugging Face | Framework and hub for interoperable reinforcement learning environments |
| OpenSandbox | Alibaba | General-purpose sandbox platform with SDKs, a sandbox protocol and Docker and Kubernetes runtimes |
The release says they were announced at AGNTCon + MCPCon North America, held October 22-23 in San Jose. It lists the existing projects as A2A, Agent Router, agentgateway, AGENTS.md, goose and MCP.
What I would check
These are my own untested suggestions.
1. Pin the spec version you build against. The roadmap mentions versioned releases but none exist yet in what I read.
2. Treat third-party skills as code. A skill can ship scripts. Read them before installing, as you would a package.
3. Test your skills in more than one agent. Portability is the pitch. Check that your SKILL.md loads and behaves the same where you need it.
4. Watch the proposal process. Once it exists, spec changes will go through it, which may change fields you rely on.
5. Look at the license and governance docs when they appear under AAIF, rather than assuming they match the earlier repository terms.
What I would not conclude
- That 70 integrations means identical behavior everywhere. The post gives no compatibility testing.
- That donation means no vendor influence. Anthropic's people are Lead Maintainers for now.
- That the security guidance will be complete or on a given date.
- That the 280 million installs figure measures active use.
Frequently asked questions
What is Agent Skills?
An open standard where a skill is a folder with a SKILL.md file and optional scripts and references, per Anthropic.
Who owns it now?
Anthropic says it is donating the name, logo, GitHub organization and agentskills.io domain to AAIF under the Linux Foundation.
Do I need to join AAIF to use it?
No. Anthropic says you can use the standard or contribute without joining.
Who maintains it?
Engineers from Anthropic, Google, OpenAI and other companies, as individuals. David Soria Parra and Mahesh Murag are the Lead Maintainers.
What is on the roadmap?
Anthropic proposed versioned releases, a formal proposal process and security guidance for authors and installers of skills.
Which other projects joined AAIF the same day?
Open Responses from OpenAI, OpenEnv from Hugging Face and OpenSandbox from Alibaba, per the Linux Foundation.
How widely is it used?
Anthropic says more than 70 platforms have integrated it and 1 million skills have been published. These are vendor figures.
Sources
- Anthropic, Donating Agent Skills to the Agentic AI Foundation (October 22, 2026): donation terms, maintainers, roadmap proposal and usage figures.
- Linux Foundation, Agentic AI Foundation Announces Four New Projects (October 22, 2026): the four projects, AAIF project count and the conference.
Related reading
- Claude Duet: Anthropic's Proactive Agent Preview and the Design Choices Builders Can Learn From
Anthropic opened a waitlist research preview of Claude Duet, an agent that checks your apps every few hours and asks approval before it sends, pays, signs or deletes. What the launch post says, and the agent design patterns worth studying.
- Anthropic Published Its Claude Unintended-Actions Report. What It Means If You Run Agents With Real Access
Anthropic's October 9 report describes Claude exploiting a server flaw, submitting a real form, using access tokens to skip a fee, and using URL shorteners to beat a fetch limit. Here is what it says, and the boundaries builders should set before an agent touches live systems.
- Anthropic's Five Claude Exploit Chains: What the Frontier Red Team Post Claims, and What It Leaves Out
Anthropic says an unreleased internal model found and chained vulnerabilities in Chrome, Safari, Android, Windows and Linux. What the October 27 post claims, the disclosure terms, the defender advice, and the numbers it does not give.
The build log
Follow the work.
New essays and prototypes, in your feed reader. Email subscriptions are paused while I connect a working delivery service. No email is collected here.
Follow via RSS