Anthropic gave Agent Skills to the Linux Foundation's AAIF. Over 70 agent platforms already read the same SKILL.md format
The format for teaching agents how your business does things now has a neutral home, and that changes how much a well-written process is worth.
What happened
Anthropic donated Agent Skills, the open standard for giving AI agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF) under the Linux Foundation. Anthropic co-founded the AAIF with Block and OpenAI and had already donated the Model Context Protocol (MCP) to it.
A skill is a folder with a SKILL.md file that teaches an agent a specific task, plus any scripts or reference material it needs. Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills since the standard was published in December 2025. So far 1 million skills have been published and nearly 280 million installed.
The donation covers the name and logo, the GitHub organization and the agentskills.io domain. Engineers from Anthropic, Google, OpenAI and other companies will maintain it as individuals. The proposed roadmap includes versioned releases, a formal proposal process and security guidance for people who write and install skills.
The spec site explains how agents load skills: at startup they read only each skill's name and description, read the full SKILL.md when a task matches, then follow it, running bundled code if needed.
My take
For operators, this is the practical side of agents. Every client has processes that live in one person's head: how a lead is qualified, how a refund is handled, what goes into a weekly report. Written as a skill, that knowledge becomes a versioned file any compatible agent can use, instead of a prompt pasted into one tool.
Two cautions. First, the description field decides when a skill gets used, so write it precisely. Second, the roadmap calls out security guidance for installing skills for a reason: a skill can bundle scripts. Treat a third party skill like any code you run on client systems, and read it before you install it.
More posts
- n8n's guide to catching AI hallucinations says no single check is enough. Layer them and give every flagged answer a place to goOct 23, 2026
- Gartner: only 13% of 2,002 CIOs and tech leaders report major returns from AI. The winners decide early what to shut downOct 23, 2026
- 1Password now lets AI agents sign in without seeing your password, as major US banks restrict agents that log in with customer credentialsOct 23, 2026
