Standard transferred to Linux Foundation affiliate
Anthropic has donated the Agent Skills standard to the Agentic AI Foundation, a neutral body operating under the Linux Foundation. Anthropic co-founded the organization last year with Block and OpenAI, previously using the foundation to host the Model Context Protocol. The transfer hands over ownership of the project brand, the agentskills.io domain, and the central GitHub organization hosting the specification.
Engineers from Anthropic, Google, and OpenAI will maintain the standard as individual contributors rather than corporate delegates. David Soria Parra and Mahesh Murag will serve as the project lead maintainers. Outside developers do not need to join the foundation to submit changes or use the standard.
Folder format and platform adoption
The Agent Skills specification defines a standard folder structure containing a single SKILL.md file. This document explains workflows and domain instructions to an agent, accompanied by local execution scripts and reference material. Anthropic created the layout during work on Claude Code to supply procedural knowledge that general language models regularly lack when tackling practical work.
Anthropic reports that more than 70 agent platforms currently read the format, including VS Code, GitHub Copilot, Cursor, and ChatGPT. The company states that the wider developer community has published roughly one million skills and recorded nearly 280 million installations since the specification launched in December 2025.
Governance roadmap and security
The new foundation roadmap prioritizes versioned specification releases, a formal change proposal workflow, and explicit security guidance. Because skills often bundle operational scripts and automation steps directly into agent tools, the project plans specific safety advice for people writing and installing skills.
Our assessment
Neutral governance separates skill authoring formats from any single model vendor, lowering the risk that custom internal workflows become locked to proprietary platforms. Engineering teams packaging procedural skills must nevertheless inspect how their chosen host tools execute local scripts referenced in SKILL.md files. Because community repositories have already logged millions of downloads, teams should implement internal signing and vetting pipelines before allowing autonomous agents to pull community skills into secure development environments.
Source: Anthropic original publication.
Read the original material
Anthropic (opens in a new tab)The article date belongs to Agentic Horizon. The source publication date is listed separately. Vendor claims remain attributed; we have not independently tested the reported capability.
