Launches
Agent Skills move to the Agentic AI Foundation: what SKILL.md files mean for small businesses
Anthropic gave Agent Skills, the SKILL.md standard used by 70+ AI agent platforms, to the Agentic AI Foundation. What changes and how to use skills safely.
Key takeaways
- On October 22, 2026, Anthropic donated Agent Skills, the open SKILL.md standard for teaching AI agents a task, to the Agentic AI Foundation under the Linux Foundation.
- Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, support skills, with 1 million skills published and nearly 280 million installs.
- A skill is just a folder with a SKILL.md file: a short name and description, then plain-language steps, plus optional scripts and reference files.
- Neutral governance makes it safer to invest in writing your own skills, because one skill file can travel between agent tools.
- Skills can also carry risk: a study of 98,380 public skills confirmed 157 malicious ones, so read every skill before you install it.
Agent Skills, the open format that teaches AI agents how to do a specific job, now has a neutral owner. On October 22, 2026, Anthropic donated the standard to the Agentic AI Foundation (AAIF), the Linux Foundation group it co-founded with Block and OpenAI, and the same home it gave the Model Context Protocol (MCP) (Anthropic). The news trended on X, where our count of posts mentioning "agent skills" or the foundation rose from roughly 280 to 500 a day earlier in the week to 1,189 on October 22 and 1,572 on October 23 (UTC days).
If you use ChatGPT, Claude, Copilot or Cursor for business work, this matters more than it sounds. Skills are the simplest way to stop re-explaining your process to an AI every time, and a neutral standard means the skills you write today should keep working when you switch tools.
What happened on October 22
Anthropic handed over the standard
Anthropic says it is donating the Agent Skills name and logo, the GitHub organization and the agentskills.io domain. You can keep using the standard, or contribute to it, without joining the AAIF. Engineers from Anthropic, Google, OpenAI and other companies will maintain it as individuals, with David Soria Parra and Mahesh Murag as lead maintainers (Anthropic).
Three other projects joined at the same time
The foundation announced four new projects at AGNTCon + MCPCon North America in San Jose: Agent Skills from Anthropic, Open Responses from OpenAI, OpenEnv from Hugging Face and OpenSandbox from Alibaba. That brings the AAIF to 10 hosted projects, up from three when it launched in December 2025 (AAIF). SiliconANGLE describes Open Responses as a shared template for model APIs, so code can move between AI providers with fewer changes (SiliconANGLE).
How big skills already are
Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have added support, and that 1 million skills have been published and nearly 280 million installed (Anthropic). The foundation's blog adds that the specification is open source under Apache 2.0 and CC BY 4.0, and that skills are used for software, marketing, design, writing and research work (AAIF blog).
What an agent skill actually is
A folder with one required file
A skill is a folder that contains a SKILL.md file, plus any optional scripts, reference documents or templates. The file starts with two required fields: a name (up to 64 lowercase letters, numbers and hyphens, matching the folder name) and a description of up to 1,024 characters that says what the skill does and when to use it. Everything after that is plain Markdown instructions (Agent Skills specification).
Why skills stay cheap to run
Agents load skills in stages. Only the name and description, about 100 tokens, are read at startup. The full instructions load when a task calls for the skill, and the spec recommends keeping them under 5,000 tokens and 500 lines, with longer material moved into separate reference files (Agent Skills specification). That is why you can keep dozens of skills installed without slowing every chat down.
How real teams use them
The AAIF blog gives examples. At Brainlabs, employees submit skills through Notion, and a person reviews each one before it goes out. At Block, engineers share an internal marketplace of hundreds of skills, used with Goose, Codex, Claude and Amp (AAIF blog).
The security catch
Malicious skills already exist
A skill can include scripts that run with your permissions, so installing one is closer to installing software than to copying a prompt. Researchers who studied 98,380 skills from two community registries, collected in January 2026, confirmed 157 malicious ones containing 632 vulnerabilities. More than half came from a single actor impersonating brands, 84.2% of the problems sat in the SKILL.md text rather than in code, and the registries removed all 157 after the researchers reported them (arXiv).
What the vendors tell you to check
Anthropic's enterprise guide warns that scripts in a skill run "with full environment access" and lists what to look for before installing: bundled scripts, instructions to hide actions from the user, web addresses or network calls, and hard-coded passwords or keys. It says to "treat Skill installation with the same rigor as installing software on production systems" (Claude docs). Anthropic says the new project roadmap will include security guidance for people who write and install skills (Anthropic).
What we did not verify
We did not test skills on all 70-plus platforms, and support for optional fields can vary; the spec marks the allowed-tools field as experimental. The adoption numbers come from Anthropic and the AAIF and were not independently checked. The malware study covers two specific registries as of January 2026, so the share of bad skills elsewhere today may be different.
What this means for small businesses
Skills turn the way you already work into something any AI agent can follow. Here is a practical way to start:
- Pick one repeated task. Choose something you explain to an AI at least weekly, such as quoting a job or answering a refund request. The AI time savings calculator helps you rank which task is worth it.
- Write the steps down first. Use our prompt to write an SOP from a messy process, or the guide on AI prompts for SOPs, reports and policies.
- Turn it into a SKILL.md. Give it a clear name and a description that says when to use it, then paste your steps. Check the instructions with the prompt grader or tighten them with the AI prompt generator.
- Keep scripts out at first. A text-only skill is easier to review. Add code only when you understand what it does.
- Vet every skill you download. Read the whole folder, look for web addresses, password requests or "don't tell the user" wording, and skip anything you can't explain.
- Keep one master copy. Store your skills in one shared folder so a fix reaches everyone, the way Brainlabs and Block do.
Frequently asked questions
What are agent skills?
Agent skills are folders of instructions, and optional scripts and resources, that teach an AI agent a specific task. Anthropic created the format and donated it to the Agentic AI Foundation on October 22, 2026 (Anthropic).
What is a SKILL.md file?
It is the one required file in a skill. It opens with a name and a description that says when to use the skill, followed by Markdown instructions (Agent Skills specification).
Do agent skills work in ChatGPT and GitHub Copilot?
Anthropic says more than 70 agent platforms support the standard, including ChatGPT, Cursor, GitHub Copilot and VS Code (Anthropic). Support for some optional fields varies between tools (Agent Skills specification).
Are agent skills safe to install?
Not automatically. A study of 98,380 public skills confirmed 157 malicious ones (arXiv), and Anthropic advises reviewing every file and script before installing a skill (Claude docs).
What is the Agentic AI Foundation?
The AAIF is a Linux Foundation group that hosts open agent standards, including MCP, AGENTS.md, goose and now Agent Skills. It reached 10 projects on October 22, 2026 (AAIF).
Sources
- Anthropic: Donating Agent Skills to the Agentic AI Foundation
- Agentic AI Foundation: AAIF announces four new projects
- Agentic AI Foundation blog: Agent Skills, sharing team expertise across AI agents
- Agent Skills: Specification
- SiliconANGLE: Agentic AI Foundation adds four more open-source AI tools
- arXiv: Detecting and understanding malicious agent skills in the wild
- Claude docs: Skills for enterprise
- X trending: Anthropic donates Agent Skills standard to neutral foundation
How we report: every fact, number and quote above comes from the linked sources, which we read before publishing. Product details can change quickly during previews, so check the vendor page before you buy. Spotted an error? Tell us via the link on our home page.