Anthropic Hands Agent Skills to the Agentic AI Foundation

Anthropic announced on October 22 that it is donating the Agent Skills standard to the Agentic AI Foundation (AAIF), the Linux Foundation body that already hosts the Model Context Protocol. The name, logo, GitHub organization and agentskills.io domain all change hands. Ten months after launch, the format claims more than 70 integrated platforms and close to 280 million installs.

#A Standard Already Widely Adopted

A skill is simply a folder containing a SKILL.md file describing a task or domain, plus any scripts or reference material it needs. Launched in October 2025 and published as an open standard in December of that year, the format quickly spread well beyond the Claude Code ecosystem: ChatGPT, Cursor, GitHub Copilot and VS Code are among the platforms that have integrated it. Anthropic puts the number of published skills at roughly one million.

#Why Anthropic Is Letting Go

The format grew out of an operational observation made while building Claude Code: models handle general knowledge work well, but real business tasks need procedural knowledge that a prompt alone doesn’t convey effectively. Rather than keep Agent Skills as a proprietary edge, Anthropic chose to publish the standard openly from the start — the same choice it made for the Model Context Protocol, donated to the same foundation. The Agent Skills donation extends that logic: a format that has become shared infrastructure across competing vendors can’t stay governed by just one of them without slowing its own adoption.

#What Changes in Governance

Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard, acting as individuals rather than on behalf of their employers. David Soria Parra and Mahesh Murag become lead maintainers. It remains possible to use or contribute to the standard without joining AAIF. A roadmap has been proposed: versioned releases, a formal change-proposal process, and security guidance for anyone writing or installing skills — a sensitive point, since a skill can bundle scripts an agent will execute.

#What a Skill Looks Like

The minimal structure is deliberately simple: a Markdown file with a YAML header describing the skill, followed by natural-language instructions.

---
name: laravel-code-review
description: Checks migrations and validation rules before merging
---
# Laravel Code Review

1. Flag any non-reversible migrations.
2. Verify every FormRequest validates the expected types.
3. Report N+1 queries in Eloquent relationships.

That simplicity helps explain the format’s fast adoption across agents as different as coding assistants and editorial agents running in production. The format has even spread beyond Claude Code into development kits such as the Laravel AI SDK.

A format published ten months ago already claims more than 280 million installs and governance shared between direct competitors — an unusually fast adoption curve for such a young standard.

#What to Remember

• Agent Skills joins the Model Context Protocol under the governance of the Agentic AI Foundation, hosted by the Linux Foundation.
• The standard claims more than 70 integrated platforms, including direct competitors of Anthropic.
• Maintenance now falls to engineers from several companies, acting as individuals.
• Security guidance for writing and installing skills has been announced but not yet published.

#Go Further

Full announcement: Anthropic, “Donating Agent Skills to the Agentic AI Foundation”. Related guide: AI for web developers.