Agent Skills Can Now Run Your Business Playbook On Any AI Tool

Anthropic has donated Agent Skills, the open format that lets an AI agent load step by step instructions for a specific task, to the Agentic AI Foundation under the Linux Foundation. More than 70 platforms, including ChatGPT, Cursor, GitHub Copilot and Claude, already use it, so a skill you write once now runs everywhere.

What did Anthropic announce?

On 22 October 2026, Anthropic handed over the Agent Skills name, logo, GitHub organisation and the agentskills.io domain to the Agentic AI Foundation (AAIF), a neutral body under the Linux Foundation that Anthropic co-founded with Block and OpenAI. The AAIF already stewards the Model Context Protocol (MCP), the connector standard that lets AI agents reach outside tools and data. Agent Skills now joins it: one gives agents tools to use, the other gives them the know-how to use them well.

Like MCP, it started as something we built at Anthropic, and I'm glad so many others have found it useful. And like MCP, it's time for it to be stewarded by the Agentic AI Foundation, where it stays open and community-driven.from David Soria Parra, Lead Maintainer of Agent Skills, Anthropic (LinkedIn, 22 October 2026)

What is an Agent Skill, and what can it actually do?

A skill is a folder containing a SKILL.md file that teaches an agent a specific task or domain, plus any scripts or reference material it needs. Anthropic built the idea inside Claude Code: the model handled general knowledge work well on its own, but it often lacked the procedural know-how that real business tasks need, the exact tone for a client email, the format a finance team expects, the steps a specific report always follows.

• Launched in October 2025 and published as an open standard two months later, Agent Skills has reached 1 million published skills and nearly 280 million installs.
• More than 70 agent platforms now load the same skill folder, among them ChatGPT, Cursor, GitHub Copilot, VS Code and Claude.
• David Soria Parra and Mahesh Murag stay on as Lead Maintainers, joined by engineers from Anthropic, Google, Microsoft and OpenAI.
• The roadmap ahead covers versioned releases, a formal proposal process, and security guidance for writing and installing skills safely.

How could your business use Agent Skills?

Picture a clinic that writes one skill for drafting patient recall messages in its own tone and WhatsApp format. Load that skill into Claude, and if the front desk also uses ChatGPT for other tasks, the same instructions work there too, so the process is trained once instead of rewritten as a fresh prompt for every tool. A retailer could build a skill for answering stock and pricing questions the way its own staff would, and an accounting firm could package exactly how it wants a monthly statement summarised. Each skill is reusable for every new hire and every AI tool already on the desk, which is the real saving: less retraining, less prompt rewriting, more consistent output.

Also today

• Anthropic made its own Claude models directly available through Google Cloud Marketplace, with Google Cloud IAM controlling access and a new Fast mode for Claude Opus 5.5 (Claude Platform release notes, 22 October 2026).
• Malaysia's The Star reported that technology leaders remain divided over whether AI development needs to slow, as governments face mounting pressure to introduce tougher safeguards (The Star, 24 October 2026).

What should you try this week?

If your team already uses Claude or ChatGPT for a repetitive task, write that process down once as a short, plain instruction file rather than retyping a prompt every time, that is already most of a skill. Because the format is now an open standard stewarded outside any single company, it is a safe thing to invest an afternoon in: it will keep working even as your team tries other AI tools. We walk through exactly this kind of reusable setup, alongside Claude Projects, Cowork and MCP connectors, in our Kota Kinabalu classes for AI Tool Mastery for Work Productivity.