Summary: On October 22, 2026, Anthropic donated the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation, the same neutral home that already hosts MCP. The name, logo, GitHub organization, and agentskills.io domain all move to the foundation. If you build agents or publish skills, the format stays the same — but its roadmap now runs through community governance instead of one vendor.
Why the handoff matters for builders
Agent Skills started as an internal tool at Anthropic while the team built Claude Code: general-purpose agents kept missing the procedural knowledge real business tasks need. A skill is deliberately simple — a folder with a SKILL.md file that teaches an agent a task or domain, plus any scripts or reference material it requires.
The standard has grown well past one company’s control. Since Anthropic published it in December 2025, more than 70 agent platforms including ChatGPT, Cursor, GitHub Copilot, and VS Code have integrated it. The Anthropic announcement reports 1 million published skills and nearly 280 million installs. When a format that many platforms depend on lives in one vendor’s repo, every downstream builder carries platform risk. Moving it to a foundation removes that.
What actually transfers
The donation covers the Agent Skills name and logo, the GitHub organization, and the agentskills.io domain. Engineers from Anthropic, Google, OpenAI, and other companies will maintain the standard as individuals, with David Soria Parra and Mahesh Murag as Lead Maintainers.
Two practical points stand out. First, you can keep using or contributing to the standard without joining the AAIF — nothing about the license or workflow changes today. Second, this mirrors the path MCP already took into the same foundation, so there is a working precedent for how the governance will feel in practice.
The proposed roadmap
Anthropic has proposed starting with versioned releases, a formal proposal process, and security guidance for people who write and install skills. That last item matters most if you distribute skills to users: installing someone else’s skill means letting untrusted instructions and scripts near your agent, and shared security guidance has been missing. Feedback on specification changes happens through the project’s Discord and GitHub.
How to prepare
If you already publish skills, audit your folder structure against the current agentskills.io spec now, before versioned releases introduce formal changes. If you’re writing your first skill, the pattern is small and procedural — teach one task well. That’s the same principle behind giving coding agents deployment-specific facts rather than generic context, which I covered in Give Your Coding Agent Deployment Facts, Not Just a Model Name.
The grounded takeaway: the skills format you rely on is now governed like infrastructure, not like a product feature. Watch the proposal process — the decisions about versioning and security guidance will define how safely skills can be shared at scale.
Sources
View the original post on X