Agent Skills, Open Responses, OpenEnv and OpenSandbox Join the Agentic AI Foundation
At AGNTCon in San Jose, the Linux Foundation project took in four projects from Anthropic, OpenAI, Hugging Face and Alibaba, bringing the foundation to ten.
Outspoken Digest Technology Desk
Friday, October 23, 2026/2 min read
Four rival AI companies handed the same neutral home a piece of their agent plumbing on Thursday. The Agentic AI Foundation, a Linux Foundation project, announced on 22 October in San Jose, timed to the AGNTCon and MCPCon conference, that it is adding Agent Skills, Open Responses, OpenEnv and OpenSandbox, according to the foundation's release.
The foundation now hosts ten projects, up from three at its launch in December 2025. Its chief technology officer and co-founder, Manik Surtani, framed it as a neutral place for shared standards and infrastructure.
What are the four new projects?
Agent Skills comes from Anthropic: a portable format for packaging instructions, scripts and resources that an agent loads when it needs them. Open Responses comes from OpenAI: a specification for multi-provider model interfaces, based on its Responses API. OpenEnv, from Hugging Face, is a framework and hub for reinforcement learning environments built on containerised services, and the release lists Microsoft, Meta-PyTorch, Reflection, Modal and Prime Intellect among participants. OpenSandbox, from Alibaba, is a general-purpose sandbox platform with multi-language SDKs and Docker and Kubernetes runtimes.
The release quotes David Soria Parra, co-creator and lead maintainer of the Model Context Protocol, saying Anthropic is donating Agent Skills so the community can shape its direction. Alibaba's Xiaobin Xu said "we don't want its future to be defined by a single company" about OpenSandbox.
What else does the foundation already host?
The existing projects are Agent2Agent (A2A), Agent Router, agentgateway, AGENTS.md, goose and the Model Context Protocol. The foundation's own site also lists eight working groups covering accuracy and reliability, agentic commerce, governance and regulatory alignment, identity and trust, observability, security and privacy, workflow integration, and a shared taxonomy, as its homepage shows. The conference itself runs 22 and 23 October at the San Jose McEnery Convention Center.
Why does neutral governance matter for agents?
Agents become more useful when tools, skills and sandboxes work across vendors, and less so when each lab's format is a private moat. Moving a specification into a foundation does not guarantee that rivals will adopt it, but it lets outside contributors vote on its direction and removes the donor's ability to change it unilaterally.
The release does not give adoption figures for the new projects, and it does not say how the foundation will decide between overlapping specifications. Those governance choices, more than the donations, will show whether ten projects add up to one coherent stack by next year's conference.
Published in The Outspoken Digest
Editorial desk
Outspoken Digest Technology DeskSoftware, hardware, artificial intelligence and what they change for everyone else.
Newsletter
The Digest, in your inbox
One edition, sent when it is ready. No noise, and your address is never passed on.
Read Next
More Technology →
NASA's Dragonfly Flies Its Hardest Titan Manoeuvre, Autonomously, in Its Most Thorough Simulation Yet
Oct 26, 2026/2 min read
Quest Diagnostics Confirms Hackers Reached Patient Records Through Its Quanum Lab Platform
Oct 26, 2026/2 min read
Instacart Is Reportedly Exploring a Sale, With DoorDash and Uber Said to Have Held Talks
Oct 26, 2026/2 min read
NASA Ends the Swift Mission After Nearly 22 Years as Its Orbit Decays and Re-Entry Looms
Oct 25, 2026/2 min read