Agent Skills moves to a shared foundation as portable AI workflows mature

Anthropic has donated Agent Skills to the Agentic AI Foundation, giving the open workflow format a vendor-neutral home.

Anthropic announced on 22 October that Agent Skills is moving to the Agentic AI Foundation under the Linux Foundation. The donation covers the standard's name and logo, GitHub organisation and agentskills.io domain. It puts an increasingly common building block for AI workflows into a shared governance structure.

According to Anthropic, more than 70 platforms have integrated the format, including ChatGPT, Cursor, GitHub Copilot and VS Code. Engineers from Anthropic, Google, OpenAI and other organisations will maintain it as individual contributors, with David Soria Parra and Mahesh Murag serving as lead maintainers. Participation does not require joining the foundation.

The format behind the adoption

A skill is a directory organised around a SKILL.md file. Its metadata tells a compatible agent what the skill does; the instructions explain the procedure. Scripts, reference documents and templates can live beside that file. The official specification uses progressive disclosure: clients first load a skill's name and description, read its full instructions when the task calls for them, and then load supporting material as needed.

That organisation gives teams a practical way to package expertise. A presentation workflow might carry a brand template, a document-review skill can include a checklist, and a data-analysis skill can point to a maintained calculation procedure. The package can be version-controlled and reviewed like other project material, while compatible clients discover it through the same basic structure.

A shared roadmap

The initial roadmap proposed by Anthropic includes versioned releases, a formal proposal process and guidance for skill authors and installers. The foundation therefore takes on both the format's stewardship and the way its community develops changes.

For developers, the useful shift is portability of procedural knowledge. Teams can document how work should be completed once, keep the supporting resources nearby and bring that package into several compatible tools. The model still supplies reasoning and language capability; the skill supplies the organisation's repeatable process.

The next work for maintainers is concrete: identify the owner of each skill, keep examples aligned with the current application, and review changes alongside the scripts or templates they affect. This turns an instruction folder into a maintained project dependency and makes its evolution visible to the people who rely on it.

Sources

- [Donating Agent Skills to the Agentic AI Foundation](https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation) — Anthropic · 2026-10-22 - [Agent Skills Overview](https://agentskills.io/home) — Agent Skills