Illustration of AI agent robots exchanging skill puzzle pieces around a central neutral hub

Anthropic Gives Agent Skills a Neutral Home: Linux Foundation’s Agentic AI Foundation

ByChris Wade

Key takeaway: On October 22, 2026, Anthropic donated its Agent Skills standard to the Linux Foundation’s Agentic AI Foundation, handing over the name, logo, GitHub organization, and agentskills.io domain. Three other companies followed suit the same day — OpenAI, Hugging Face, and Alibaba — pushing the year-old foundation to 10 hosted projects and making AI agent interoperability a neutral, vendor-independent standard.

What happened on October 22

The Agentic AI Foundation (AAIF) announced four new projects at AGNTCon + MCPCon North America, held October 22–23 at the San Jose McEnery Convention Center: Agent Skills contributed by Anthropic, Open Responses contributed by OpenAI, OpenEnv contributed by Hugging Face, and OpenSandbox contributed by Alibaba, according to the Linux Foundation’s press release.

The four additions take the AAIF from three projects to ten hosted projects in its first year, joining the Agent2Agent (A2A) protocol, Agent Router, agentgateway, AGENTS.md, goose, and the Model Context Protocol (MCP). The foundation launched in December 2025 — co-founded by Anthropic, Block, and OpenAI — with three projects: MCP, goose, and AGENTS.md.

For Agent Skills, the donation covers four assets: the name, the logo, the GitHub organization, and the agentskills.io domain. Anthropic reports that more than 70 agent platforms — including ChatGPT, Cursor, GitHub Copilot, and VS Code — have integrated the format since its December 2025 publication, with one million skills published and nearly 280 million installations. Those figures are Anthropic’s own, not an independent count, as SiliconReport noted.

What an Agent Skill actually is

An Agent Skill is a folder. Inside it sits a SKILL.md file with metadata and instructions, plus optional folders for scripts, reference documents, and assets. When an AI agent faces a task the skill describes — filling in a company spreadsheet, reviewing a contract against a firm’s internal rules, extracting tables from PDFs — it reads the skill and loads the know-how at that moment.

The idea grew out of Anthropic’s internal engineering on Claude Code, which launched in October 2025. The team found it lacked procedural knowledge for real business tasks, so they built skills to teach agents specific workflows. Within weeks, developers had published thousands of skills and built loaders for Codex, Cursor, and OpenCode — community uptake that pushed Anthropic to publish the format as an open standard in December 2025 so any agent could use it.

The four new AAIF projects cover four layers of the agentic stack, per the Linux Foundation: how agents gain new skills (Agent Skills), how applications talk to models across providers (Open Responses), how agents are trained and evaluated (OpenEnv), and where agents run safely (OpenSandbox).

Why this matters

AI agents are becoming the next software battleground, and the risk everyone wants to avoid is lock-in: if every company builds its own incompatible way for agents to use tools and share workflows, a skill written for one assistant is useless in another, and switching tools means starting over.

Shared, neutrally governed standards fix that. Manik Surtani, CTO and co-founder of the Agentic AI Foundation, put it plainly in the announcement: the mission is giving the community “one place to build shared standards, protocols, and infrastructure, with no single company controlling the outcome.” It is also a repeat of a playbook that already worked: Anthropic donated MCP to the same foundation in December 2025, and MCP became the industry’s default standard for connecting agents to tools.

What this means for you

If you use AI coding assistants or agents at work, nothing changes today — your SKILL.md files still work exactly as before. The change is about the future: the standard now has a vendor-neutral home, and engineers from Anthropic, Google, OpenAI, and other companies will maintain it as individuals, with David Soria Parra and Mahesh Murag as lead maintainers.

The practical takeaway for teams: build reusable skills for your own workflows rather than rebuilding instructions in every AI tool. Because the format is open and now neutrally governed, the methods you encode have the best chance of surviving the next product cycle — you can carry them with you even if you switch assistants. This connects to the broader shift toward AI agents doing real operational work, and to the ongoing debate about how much autonomy agents should have.

One caution: a neutral standard makes skills easier to share — and therefore easier to inherit from strangers. Security researchers have flagged that skills can tell an agent to run arbitrary scripts. Audit what your skills execute before adopting community-built ones.

Our analysis: governance as a growth strategy

The following is analysis, not a verified fact. Companies tend to donate a format to a foundation when they want everyone else to stop worrying about who owns it. Experiments stay in-house; donated standards are the ones meant to become infrastructure. By giving away Agent Skills at 280 million installs, Anthropic appears to be betting that a neutral standard it shaped will beat a proprietary one it has to defend.

The more interesting subplot: four rival companies chose to contribute to neutral governance on the same day — much like container standards (OCI) and cloud-native tooling (CNCF) before it. The winner of the agent era may be decided less by who has the smartest model and more by whose workflows everyone shares.

Frequently asked questions

What is an Agent Skill?

An Agent Skill is a folder containing a SKILL.md instruction file plus optional scripts, reference documents, and assets. It gives an AI agent the specialized knowledge and workflow steps needed for a specific task, and the agent loads it only when the task calls for it. The format is open and free to use.

Who owns Agent Skills now?

Nobody owns it commercially anymore. Anthropic donated the name, logo, GitHub organization, and agentskills.io domain to the Agentic AI Foundation under the Linux Foundation on October 22, 2026. Engineers from multiple companies maintain it as individuals under open, vendor-neutral governance — no foundation membership or paid account is needed to use or contribute.

Do I need to change my existing skills?

No. This is a governance change, not a format change — existing SKILL.md files keep working. But because the neutral standard makes skills easier to share, audit what your skills tell agents to run before adopting community-built ones, since a skill can include executable scripts.

How does this relate to the Model Context Protocol (MCP)?

Both are now hosted by the same foundation under the Linux Foundation. MCP, which Anthropic donated in December 2025, is the standard for connecting agents to tools and data. Agent Skills is the standard for packaging reusable workflows and know-how. Together they cover two different layers of the agentic stack: tool access and task knowledge.

What are the other three new AAIF projects?

Alongside Agent Skills, the foundation added Open Responses (contributed by OpenAI) for model interoperability, OpenEnv (Hugging Face) for agent training environments, and OpenSandbox (Alibaba) for secure execution. The announcement brings the AAIF to 10 hosted projects in its first year.

Sources


Similar Posts