What has Anthropic handed to the AAIF?
On 22 October 2026, Anthropic donated the Agent Skills name and logo, the associated GitHub organization and the agentskills.io domain to the Agentic AI Foundation. According to Anthropic, it co-founded the AAIF with Block and OpenAI last year, and it also donated the Model Context Protocol (MCP) there. The foundation is meant to give the standard a neutral home and to ensure the community has a say in where Agent Skills go next.
What matters in practice: according to Anthropic, anyone who uses the standard or wants to contribute does not need to join the AAIF. Agent Skills is maintained by engineers from Anthropic, Google, OpenAI and other companies who take part as individuals. The Lead Maintainers are David Soria Parra and Mahesh Murag. Anthropic stresses that it wants to keep Agent Skills open, community-driven and vendor-neutral.
What is an agent skill?
According to Anthropic, a skill is a folder containing a SKILL.md file that teaches an agent a specific task or domain, along with any scripts or reference material it needs. The Linux Foundation describes Agent Skills as an open format that packages instructions, scripts and resources in a reusable way. Agents load them on demand, which makes specialized know-how portable across agent products and platforms.
Skills were developed while Anthropic was working on Claude Code. It handled general knowledge work well, but often lacked the procedural knowledge that real business tasks need. Within weeks of the October 2025 launch, developers published thousands of skills on GitHub and wrote loaders for Codex, Cursor and OpenCode. In December 2025, Anthropic published Agent Skills as an open standard under its own GitHub organization and domain. According to Anthropic, 1 million skills have been published so far and nearly 280 million installed.
Which projects now belong to the AAIF?
According to the Linux Foundation press release of 22 October 2026, the AAIF has added four new projects and now hosts 10 projects in its first year; it launched in December 2025 with three. The new projects cover four layers: how agents gain new skills, how applications talk to models across providers, how agents are trained and evaluated, and where agents run safely. They were announced at AGNTCon + MCPCon North America on 22 and 23 October in San Jose.
Projects already in the AAIF are the Agent2Agent (A2A) protocol, Agent Router, agentgateway, AGENTS.md, goose and MCP. All projects operate under the foundation's open, vendor-neutral governance.
- Agent Skills (Anthropic): open format for giving agents new capabilities and domain expertise
- Open Responses (OpenAI): open specification for multi-provider LLM interfaces
- OpenEnv (Hugging Face): framework and community hub for reinforcement learning environments
- OpenSandbox (Alibaba): sandbox platform for AI applications with SDKs and Docker and Kubernetes runtimes
What does Open Responses offer for multiple model providers?
According to the Linux Foundation, Open Responses is an open specification for building multi-provider, interoperable LLM interfaces, based on the OpenAI Responses API. It defines a shared schema, consistent streaming events and extensible tooling. Developers are meant to describe requests and outputs once and run them across model providers with minimal translation.
Our assessment, not a vendor statement: together with Agent Skills, this touches two places where dependence on a single provider can arise, namely the agents' working knowledge and the technical interface to the model. Anyone pursuing a strategy with several model providers should keep an eye on both projects.
What is on the proposed roadmap?
Anthropic has proposed starting with a roadmap that covers three points. As with MCP, the project will continue to prioritize community input and transparent decision-making. Feedback on proposed specification changes can be given via the project's Discord, and the discussion also takes place on GitHub.
In the press release, Anthropic's David Soria Parra explains the donation as follows: Agent Skills was launched as an open standard so that developers and enterprises can make their agents more capable. Handing it to the AAIF is meant to ensure that anyone building with skills can shape what comes next.
- Versioned releases of the standard
- A formal proposal process
- Security guidance for people who write and install skills
What does the donation mean for businesses?
Our assessment, not a vendor statement: because Agent Skills is, according to Anthropic, already integrated into platforms such as ChatGPT, Cursor, GitHub Copilot and VS Code and originates from Claude Code, a skill written once can in principle be used in tools from different providers. Neutral governance lowers the risk that your process knowledge stays tied to a single provider. This pays off especially when teams already use different AI tools today.
At the same time, governance becomes more important. A skill can contain scripts, and nearly 280 million installations show how quickly third-party skills spread. The fact that the proposed roadmap includes security guidance for writing and installing skills underlines the topic. So define who reviews and approves third-party skills before they are installed in your company.
What should businesses do now?
Record which skills are already used in your AI tools and where they come from. Capture recurring workflows as your own skills in an internal, versioned library instead of tying them to a single tool. Review third-party skills before installing them, especially any scripts they contain, and govern approvals in your AI policy. The specification at agentskills.io is the authoritative reference.
If you use several model providers, also keep an eye on Open Responses. The free AI System Check helps organize tools, skills and approvals across your team. These notes are suggestions, not customer results.
Keep it verifiable
