Anthropic Transfers Agent Skills Project to Foundation
The open-source SKILL.md format moves to the Agentic AI Foundation to foster cross-platform standardization for AI agents.
Updated on Oct. 22, 2026 in Artificial Intelligence
Live Poll
Do you trust the security of automated AI tools downloaded to your computer?
Anthropic has transferred its Agent Skills project and the associated SKILL.md format to the Linux Foundation's Agentic AI Foundation. This open-source standard, currently supported by over 70 agent platforms, aims to formalize how AI agents execute business tasks.
Why it matters
The transition moves a critical infrastructure project from a single corporate entity to an open governance model, facilitating broader adoption and standardization across the AI ecosystem. This change addresses the need for consistent, modular instruction sets that allow agents to handle complex enterprise workflows.
The SKILL.md format enables agents to load descriptions at startup, pulling full instructions only when a task match occurs. With 70 platforms currently integrated, the ecosystem contains 1 million published skills and 280 million installations, though 969 skills have been flagged for high-impact payloads.
The players
Anthropic
An AI research and deployment company known for the Claude model family and its focus on steerable, reliable AI systems.
Agentic AI Foundation
A Linux Foundation project that establishes open-source standards and governance for AI agent interoperability and security.
OpenAI
A research organization and developer of the GPT and o-series models that co-founded the Agentic AI Foundation.
Block
A financial technology company that co-founded the Agentic AI Foundation to support open-source AI infrastructure.
The details
A skill is structured as a folder containing a SKILL.md file alongside supporting scripts or reference materials. When an AI agent initializes, it reads the skill's name and description but defers loading the full execution instructions until a specific task match is triggered. This modular architecture was developed after developers found that early iterations of Claude Code lacked the procedural knowledge required to complete reliable business-logic tasks.
Timeline
October 2025: Agent Skills project first shipped.
December 2025: Anthropic published the format as an open standard.
December 2025: Anthropic, Block, and OpenAI co-founded the Agentic AI Foundation.
October 2026: Anthropic transferred Agent Skills to the Agentic AI Foundation.
The Tech Race
The transfer marks a departure from proprietary control, placing the project under the governance of the Linux Foundation's Agentic AI Foundation. This transition scales the project alongside the foundation's total portfolio of 10 host projects and 305 member organizations.
Engineers and developers building AI agent workflows can now access the standardized SKILL.md format under open governance. Anthropic has proposed a formal roadmap for versioned releases and security guidance, which will dictate how developers integrate these skills into future agent platforms.
The takeaway
The move to the Agentic AI Foundation signals a push toward industry-wide interoperability for autonomous agent tasks. Watch for the upcoming versioned releases and new security guidance documents on the project roadmap to see how the community addresses the 969 identified malicious agent skills.
Further reading
For more developments in agent protocols and standardized frameworks, visit the Artificial Intelligence section.
Live Poll
Do you trust the security of automated AI tools downloaded to your computer?
