Skip to content

AI Compass / News

Agent Skills: Anthropic hands the skill format to the Linux Foundation

Prompting

What moving the skill format to a neutral foundation means for your checklists, templates and writing rules.

Source: Donating Agent Skills to the Agentic AI Foundation (Anthropic),

Note: I use Claude and Claude Code for my own work. Anthropic does not pay me, and the links are not affiliate links.

Technical drawing: a folder with a SKILL.md file and two subfolders for scripts and assets, three arrows lead to three agents that load the same folder. Below, a version line with four commits.

What happened: On 22 October 2026, Anthropic announced that it is donating the Agent Skills format to the Agentic AI Foundation (AAIF), a foundation under the umbrella of the Linux Foundation. Anthropic co-founded the AAIF with Block and OpenAI in 2025. According to Anthropic, the donation covers the standard, its name and logo, the GitHub organisation and the agentskills.io domain. A skill is a folder with a SKILL.md file: a name, a short description and instructions for how an AI agent (a program that works through a task in several steps on its own) handles a specific task, plus scripts and templates where needed. Anthropic says more than 70 agent platforms have integrated the format, including ChatGPT, Cursor, GitHub Copilot and VS Code. OpenAI’s own guide states that skills in ChatGPT and Codex build on this standard. Going forward, engineers from Anthropic, Google, OpenAI and other companies will maintain it, taking part as individuals.

What this means: What you capture as a skill, such as an inspection checklist, a quote template or your writing rules, is less tied to one vendor. If you switch tools, you take the folder with you. Security guidance for people who write and install skills, however, is so far only on Anthropic’s proposed roadmap.

What you can do:

  1. Store skills neutrally: Keep your skills in a folder of your own rather than only inside one vendor’s account, under version control, for example with Git (so every change stays traceable).
  2. Sharpen the description: According to agentskills.io, an agent first reads only the name and description and loads the full skill when a task matches. So state precisely what the skill is for.
  3. Cross-check: Test an important skill in a second tool before you rely on it being portable.
  4. Read third-party skills: Check any bundled scripts before installing, because the agent can run them. The card Vet third-party skills before they run shows how.

Context: Anthropic’s figures (1 million skills published, nearly 280 million installed) are vendor claims, with no information on how they were counted.

Bringing AI into your own team? That’s what my workshops are for. See the workshops