Anthropic Hands Agent Skills Standard to AAIF: 1 Million Skills, 280 Million Installs

Anthropic has transferred the open Agent Skills standard to the Agentic AI Foundation, part of the Linux Foundation. The format lets developers give an agent specialized knowledge and workflows. Since December 2025, more than 70 agent platforms have adopted it, including ChatGPT, Cursor, GitHub Copilot and VS Code. A skill is simple: a folder containing a SKILL.md file, scripts and reference materials. About 1 million skills have been published so far, and they have been installed almost 280 million times. • The name, logo, GitHub organization and the agentskills.io domain all move to the foundation • Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard • The roadmap includes versioned releases, a formal proposal process and a security guide

AI-processed from Anthropic Blog; edited by Hamidun News
Anthropic Hands Agent Skills Standard to AAIF: 1 Million Skills, 280 Million Installs
Source: Anthropic Blog. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceAnthropic Blog ↗
Editor-in-Chief
Key metric #1The name, logo, GitHub organization and the agentskills.io domain all move to the foundation
Key metric #2Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard
Key metric #3The roadmap includes versioned releases, a formal proposal process and a security guide

Anthropic has transferred the Agent Skills standard to the Agentic AI Foundation (AAIF), which operates under the Linux Foundation. Since the standard was published in December 2025, more than 70 agent platforms have integrated it, including ChatGPT, Cursor, GitHub Copilot and VS Code. The community has already released about 1 million skills, and installs have approached 280 million. The format now has a neutral owner instead of a single company.

A folder with one file instead of a heavy SDK

A skill is a folder containing a SKILL.md file. The file describes a specific task or subject area to the agent. Scripts and reference materials the task needs can sit alongside it. There is no complex protocol or separate runtime underneath, and that simplicity is what helped the format spread quickly.

Skills first appeared inside Claude Code. The agent handled general knowledge work well, but it often lacked the procedural experience that real business tasks require: how exactly this company formats a report, which steps come before a release, in what order to call tools. A skill packages that experience in a portable form.

Within a few weeks of the October 2025 launch, developers had already posted thousands of skills on GitHub. They also wrote loaders for Codex, Cursor and OpenCode, even though nobody had asked them to. Anthropic then split the specification out into a separate open standard with its own GitHub organization and domain, so that any agent could use it.

What exactly is moving to the foundation

Anthropic is handing over more than the text of the specification. The Agent Skills name, the logo, the GitHub organization and the agentskills.io domain all move to AAIF. In other words, the foundation gets control of the project's entire infrastructure.

Anthropic, Block and OpenAI founded the foundation last year. Anthropic had previously given it the Model Context Protocol (MCP) as well. The two foundational standards of the agent ecosystem now sit side by side: MCP for connecting to tools and data, and Agent Skills for knowledge and procedures.

Key terms of the deal:

  • more than 70 platforms already support the standard;
  • about 1 million published skills;
  • almost 280 million installs;
  • the standard can be used and extended without joining AAIF;
  • maintainers work as individuals, including engineers from Anthropic, Google, OpenAI and other companies;
  • lead maintainers: David Soria Parra and Mahesh Murag.

The last point matters for trust. When engineers from competing labs oversee a standard, rather than representatives of a single vendor, other market players have less reason to fear that the format will be bent in someone's favor.

Versions, a proposal process and security

To start, Anthropic proposed a roadmap with three parts: versioned releases, a formal process for submitting proposals, and security recommendations for people who write and install skills. As with MCP, the project promises to give priority to community input and to make decisions transparently. Changes to the specification can be discussed in the project's Discord and on GitHub.

Security is arguably the most practical item here. A skill can include scripts, so installing someone else's skill carries a risk close to running someone else's code and instructions inside an agent that has access to work data. With hundreds of millions of installs, common vetting rules become a necessity, not a formality.

What teams working with skills should do

If your company builds agents, treat skills as a portable asset. Knowledge about internal processes, written down in SKILL.md, works across platforms, and there is almost no vendor lock-in.

Practical steps:

  • turn recurring work procedures into separate skills and keep them in a repository next to your code;
  • before installing third-party skills, read the scripts and instructions, not just the description;
  • watch for versioned releases of the specification so you don't break compatibility;
  • if you want to influence the format, join the discussion on GitHub or in Discord; AAIF membership is not required.

The specification is available at agentskills.io.

Story

Key Questions Answered

What are the key details on A folder with one file instead of a heavy SDK?

A skill is a folder containing a SKILL.md file. The file describes a specific task or subject area to the agent. Scripts and reference materials the task needs can sit alongside it. There is no complex protocol or separate runtime underneath, and that simplicity is what helped the format spread quickly. Skills first appeared inside Claude Code. The agent handled general knowledge work well, but it often lacked the procedural experience that real business tasks require: how exactly this company formats a report, which steps come before a release, in what order to call tools.

What exactly is moving to the foundation?

Anthropic is handing over more than the text of the specification. The Agent Skills name, the logo, the GitHub organization and the agentskills.io domain all move to AAIF. In other words, the foundation gets control of the project's entire infrastructure. Anthropic, Block and OpenAI founded the foundation last year. Anthropic had previously given it the Model Context Protocol (MCP) as well. The two foundational standards of the agent ecosystem now sit side by side: MCP for connecting to tools and data, and Agent Skills for knowledge and procedures. Key terms of the deal: mo

What are the key details on Versions, a proposal process and security?

To start, Anthropic proposed a roadmap with three parts: versioned releases, a formal process for submitting proposals, and security recommendations for people who write and install skills. As with MCP, the project promises to give priority to community input and to make decisions transparently. Changes to the specification can be discussed in the project's Discord and on GitHub. Security is arguably the most practical item here. A skill can include scripts, so installing someone else's skill carries a risk close to running someone else's code and instructions inside an age

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Autonomous AI Agents

Delegate Deep Research, Coding & Workflows to an AI Agent

Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.

Launch NomadNet Agent →NomadNet AI Platform
What do you think?
Loading comments…