Agent Skills Move to the Linux Foundation's Agentic AI Foundation
Anthropic is donating the Agent Skills standard to the Agentic AI Foundation, the same neutral home that now hosts MCP. For teams building agents, the way you package capabilities no longer depends on one vendor.
By TRAGenX Desk
What happened
On October 22, 2026, Anthropic announced it is donating the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation. In its words, the company is "committed to keeping Agent Skills open, community-driven, and vendor-neutral." Anthropic co-founded AAIF with Block and OpenAI, and it has already donated the Model Context Protocol (MCP) to the same foundation.
Simon Willison, who flagged the news, noted that he saw this coming. In December 2025, when Anthropic first described skills as an "open standard", he wrote that he wouldn't be surprised to see it end up in the AAIF, the new home of the MCP specification.
Why governance matters more than the spec
A skill is a small, reusable bundle of instructions and resources that an agent loads when a task calls for it. Those bundles end up encoding real team knowledge: how you review a pull request, how you run a migration, how you format a compliance report. That knowledge is worth little if it is locked to a single vendor's runtime.
An "open standard" owned by one company is still a company decision. It can change on that company's roadmap. Moving the standard to a foundation changes who gets a say. The details of how AAIF will govern it are something to watch, and we won't guess at them here. The direction, though, is clear: neutral stewardship.
What it means for builders
- Lower lock-in risk. If you invest in writing skills for your coding agents, a vendor-neutral home makes it more plausible that they stay portable across tools.
- One foundation, two layers. MCP and Agent Skills now share a home. They solve different problems: MCP connects agents to tools and data, while skills package the know-how for using them. Having both under one roof makes alignment between them easier.
- Skills are worth treating as code. Version them, review them and test them, the same way you would any shared engineering asset.
What we'd do now
Nothing here forces a migration. The practical move is to start capturing your team's repeatable workflows in skill form, keep them in version control, and keep them small and single-purpose. If the standard evolves under AAIF, a tidy library is far easier to adapt than a pile of ad hoc prompts.
Be careful with what goes into them. A skill steers an agent's behavior, so a skill from an untrusted source deserves the same scrutiny as a dependency. Review before you install, and never put secrets inside one.
The bigger picture
Agent infrastructure is following a familiar path: a vendor proposes a convention, other vendors adopt it, and eventually it lands in a foundation. That is how interoperability tends to get settled. For anyone doing AI-assisted development, the less the plumbing depends on a single company, the more you can focus on the part that actually differs between teams: the quality of your workflows.
FAQ
Frequently asked questions
- What is the Agentic AI Foundation (AAIF)?
- AAIF is a foundation under the Linux Foundation that Anthropic co-founded with Block and OpenAI. It is the home of the Model Context Protocol, and Anthropic is now donating the Agent Skills standard to it as well.
- What did Anthropic announce about Agent Skills?
- Anthropic said it is donating the Agent Skills standard to AAIF, and that it is committed to keeping Agent Skills open, community-driven and vendor-neutral.
- Do I need to change anything in my agent setup?
- Not based on this announcement. It changes governance, not how skills work today. It is a good prompt to organise your team's workflows as versioned, reviewable skills.
Sources