Anthropic is handing ownership of Agent Skills — the open standard that teaches AI agents specialized tasks and workflows — to the Agentic AI Foundation (AAIF), the Linux Foundation project it co-founded with Block and OpenAI. The donation, announced on October 22, transfers the Agent Skills name and logo, its GitHub organization, and the agentskills.io domain into neutral, community-driven governance.
The move follows a year of unusually fast, organic adoption. Since Anthropic published Agent Skills as an open standard in December 2025, more than 70 agent platforms — including ChatGPT, Cursor, GitHub Copilot, and VS Code — have integrated the format, according to Anthropic. Roughly one million skills have been published to date, and nearly 280 million have been installed. For more context on this story, see our ongoing AI news coverage.
What Agent Skills Actually Are
A skill is deliberately simple: a folder containing a SKILL.md file that teaches an agent a specific task or domain, along with any scripts or reference material the job requires. Anthropic developed the format while working on Claude Code, after finding that the agent handled general knowledge work well but often lacked the procedural knowledge that real business tasks need.
Adoption outpaced the company's own expectations. Within weeks of the October 2025 launch, developers had published thousands of skills on GitHub and written loaders so that competing agents — Codex, Cursor, and OpenCode among them — could read the same files. Anthropic responded by publishing Agent Skills as a formal open standard under its own GitHub organization and domain, so any agent, not just Claude, could use it.
Why Donate It Now
In its announcement, Anthropic framed the transfer as a commitment to keeping Agent Skills open, community-driven, and vendor-neutral. The AAIF will provide a neutral home for the standard going forward, ensuring the community has a say in where it goes next.
The foundation is a familiar destination. Anthropic co-founded the AAIF with Block and OpenAI last year and has already donated the Model Context Protocol (MCP), the agent-to-tool connectivity standard, to the same organization. Placing both a connectivity layer and a knowledge layer under foundation governance gives the fast-moving agent ecosystem two of its core interoperability standards on neutral ground.
What Changes for Developers
Day to day, very little. Developers can continue using the standard or contributing to it without joining the AAIF, and the specification itself does not change with the transfer.
Governance, however, becomes more formal. Engineers from Anthropic, Google, OpenAI, and other companies will maintain the standard as individuals rather than as corporate representatives, with David Soria Parra and Mahesh Murag serving as lead maintainers. Anthropic has proposed an initial roadmap covering versioned releases, a formal proposal process for specification changes, and security guidance for people who write and install skills — a recognition that a million community-published skill files are also a potential attack surface. As with MCP, the project says it will prioritize community input and transparent decision-making, with feedback routed through the project's Discord.
Interoperability Through Simplicity
The design philosophy behind Agent Skills is worth noting precisely because it resists complexity. Rather than defining a new API or runtime, the standard leans on the file formats developers already use: markdown for the instructions, ordinary scripts for the executable parts. That choice lowered the barrier to entry enough that a skill ecosystem could bootstrap itself without a central registry or an SDK release cycle, and it is a large part of why loaders appeared for competing agents within weeks rather than quarters.
It also explains why the donation is credible. A standard this simple is hard to fork in any meaningful way, since the value lies in the million published skills rather than in proprietary plumbing. Moving it to a foundation mostly changes who convenes the conversation about its future — and, in the AI industry of 2026, that convening power is not a small thing.
Standards as Neutral Ground
The donation is the second time Anthropic has placed a widely adopted agent technology into independent stewardship, and it reflects a broader pattern in the industry. Interoperability layers are most valuable when every vendor can build on them without asking permission, and a standard owned outright by one lab carries adoption risk that a foundation-governed one does not.
There is a competitive logic at work as well. With OpenAI, Google, and Anthropic each shipping capable agent platforms, the battle is increasingly over whose ecosystem collects the developers. Standards like Agent Skills and MCP make the underlying components interchangeable, which shifts competition toward the agents themselves — and makes Anthropic's research preview releases, such as the proactive-agent app Claude Duet, the places where differentiation actually happens.
For the developers who built the ecosystem's million skills, the immediate practical effect is stability: the folders they publish today will keep working, and the rules for how the format evolves will now be written in the open.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
