Anthropic Donates the Agent Skills Standard to the Linux Foundation's AAIF

Huo Xing Cai Jing reports that Anthropic has given its Agent Skills standard to the Agentic AI Foundation (AAIF), which is part of the Linux Foundation. Anthropic first created this format while building Claude Code, which launched in October 2025. In December, Anthropic made the format an open standard that different AI platforms could use. More than 70 AI agent platforms, including ChatGPT, Cursor, and GitHub Copilot, already support it. Anthropic explains that Skills are reusable sets of instructions that tell AI agents how to do tasks. Developers can write the steps for a task in files called SKILL.md. They can also include scripts and helpful reference materials. Different AI agents can then use these instructions whenever they need them. MCP helps AI agents connect to outside tools and information, while Skills explain how agents should complete specific tasks. Anthropic's donation includes the standard's name, logo, GitHub organization, and official website domain. Anthropic, OpenAI, and Block started AAIF in 2025. The foundation had already accepted MCP before receiving the Skills standard. From now on, engineers from Anthropic, OpenAI, and Google will work together to maintain the Skills standard as individuals, rather than as official representatives of their companies. This means Anthropic will no longer be the only organization leading the standard. According to Vercel, the skills.sh directory now has about one million skills, which have been installed nearly 280 million times in total. Snyk previously found 76 harmful skills among the skill samples in ClawHub. To help make the standard safer and better organized, Anthropic has suggested adding version control to track changes, a formal process for suggesting improvements, and security guidelines.