Agent Skills Joins the Agentic AI Foundation, With Open Responses
Agentic AIMore than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have adopted Agent Skills since Anthropic published the format as an open standard in December 2025. Now Anthropic is handing it to the Agentic AI Foundation, and OpenAI did the same with its Open Responses API spec on the same day. For teams writing skills or wiring agents to several model providers, the formats they depend on are moving out of any one vendor’s control.
What Happened
Anthropic’s announcement says it is donating the Agent Skills name and logo, GitHub organization and the agentskills.io domain to the AAIF, part of the Linux Foundation, which Anthropic co-founded with Block and OpenAI last year and to which it gave the Model Context Protocol. Anthropic reports 1 million skills published and nearly 280 million installs. Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals, with David Soria Parra and Mahesh Murag as lead maintainers. Developers can keep using or contributing to the standard without joining the AAIF.
The New Stack reports that OpenAI is donating Open Responses, a spec for core API calls that work across LLM providers, and that OpenSandbox and a Hugging Face framework for agent environments are joining too. They sit alongside existing AAIF projects including MCP, A2A, AGENTS.md, goose and agentgateway.
Why It Matters
A skill is simple: a folder with a SKILL.md file of metadata and instructions, plus optional scripts and reference material. The Agent Skills overview describes progressive disclosure. Agents load only each skill’s name and description at startup, read the full instructions when a task matches, then run any bundled code. That keeps many skills available with a small context footprint. Adoption was fast: Anthropic says developers published thousands of skills on GitHub and wrote loaders for Codex, Cursor and OpenCode within weeks of the October 2025 launch.
The proposed roadmap covers versioned releases, a formal proposal process, and security guidance for people who write and install skills. That last item matters: skills can bundle executable scripts, and a format installed nearly 280 million times needs shared rules for trusting them.
Open Responses targets a different layer. OpenAI says it gives developers a shared format across endpoints that support it, and that Vercel AI Gateway and Databricks Model Serving already implement it. The New Stack lists AWS Bedrock, Ollama, Hugging Face, Nvidia, OpenRouter and vLLM among supporters.
What This Means for Agent Builders
Simon Willison notes he predicted the move last December, when Anthropic first called skills an open standard. With MCP for tools, AGENTS.md for repo instructions and now skills for procedures and Open Responses for model calls, the AAIF holds much of the standard agent stack. Stack Archive’s guide to AGENTS.md vs skills covers when to use each.
The So What
Teams with internal skills should write them to the agentskills.io specification rather than one product’s variant, so they keep working across harnesses, and follow the security guidance once it’s published. Teams maintaining provider-specific integrations should check whether their model endpoints support Open Responses before writing another adapter.
Source: Anthropic
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →