Agent Skills Moves to a Neutral Foundation: What It Means for Businesses

On October 22, 2026, Anthropic announced that it is donating Agent Skills to the Agentic AI Foundation, known as the AAIF. Agent Skills is an open standard that gives AI agents specialised knowledge and step-by-step workflows. The donation covers the standard itself, the Agent Skills name and logo, its GitHub organisation and the agentskills.io domain. Anthropic says it wants to keep the standard open, community-driven and vendor-neutral.

The AAIF operates under the Linux Foundation. Anthropic co-founded it with Block and OpenAI, and it is also where Anthropic placed the Model Context Protocol, or MCP, the standard many AI tools use to connect to business software. Agent Skills now sits beside MCP in the same neutral home. Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals, and nobody has to join the foundation to keep using it or to contribute.

So what is a skill? In plain terms, it is a folder of instructions. At its centre is a single file called SKILL.md, and the folder can also hold scripts and reference material. A skill teaches an AI agent how to do one specific job the way your business does it, such as preparing a quote, checking a document against a checklist or writing a follow-up in your house style. Anthropic built the idea while working on Claude Code, after finding that agents handled general knowledge well but often lacked the procedural know-how that real business tasks need.

The standard has spread quickly. Anthropic launched skills in October 2025 and published the open standard in December 2025. According to Anthropic, more than 70 agent platforms have since integrated it, including ChatGPT, Cursor, GitHub Copilot and VS Code. The proposed roadmap under the foundation includes versioned releases, a formal process for proposing changes, and security guidance for people who write and install skills.

Why should a business owner care about a governance announcement? Because it lowers a real risk: being locked into one AI vendor. When your company's know-how lives inside one provider's proprietary format, switching providers means rebuilding it. A skill written to an open, foundation-governed standard is a plain folder of text and scripts that many platforms can read. If a better or cheaper model arrives next year, your playbooks can move with you.

It also changes where the value sits. The model is becoming the replaceable part. The lasting asset is your process knowledge: how you qualify a lead, how you answer a refund request, which details a clinic must confirm before booking. Writing that knowledge down once, in a portable format, is a sensible investment whichever AI provider you use.

Consider a dental or physiotherapy clinic in Delhi or Dubai. Front-desk staff follow unwritten rules: which treatments need a longer slot, what to ask new patients, when to send a case to the doctor. Those rules can be written as a skill and given to an AI assistant that drafts appointment confirmations, prepares intake summaries or answers routine questions. Because the skill is just a documented procedure, the clinic owner can read it, approve it and update it without a developer.

A real estate agency in London or Mumbai is another good fit. Teams spend hours turning property details into listings, comparing documents and sending follow-ups after site visits. A set of skills can capture the agency's listing format, its document checklist and its follow-up rules, so an AI agent produces consistent drafts for a human to review. An e-commerce brand in the USA could do the same for product descriptions, return-policy replies and supplier emails, keeping the brand voice consistent across every tool that supports the standard.

There are honest limits. First, security. A skill can include scripts, which means installing one is closer to installing software than reading a document. Skills downloaded from unknown sources could contain harmful instructions or code. Anthropic itself lists security guidance as a roadmap item, which tells you the guidance is still being written. Businesses should only use skills they have written or carefully reviewed.

Second, portability is improving but is not perfect. Platforms that support the standard may still differ in how they load skills, what scripts they can run and which tools they connect to. A skill that works smoothly in one product may need adjustments in another. Third, a skill only helps if the process behind it is clear. If your team does not agree on how a task should be done, an AI agent will not settle that for you. And for anything involving patient data, payments or legal documents, a human should still review the output.

This is where Legacy Services can help. We start by mapping the repetitive tasks in your business and writing down how your best people actually do them. We then turn those procedures into clean, reviewed skills that you own outright, stored in your own repository rather than locked inside a vendor account. We connect them to the systems you already use through MCP or n8n workflows, so an agent can read your CRM, update a Google Sheet or post to WhatsApp as part of the job.

In practice, that might be a custom AI agent that drafts property listings from your internal format, a website chatbot that answers customers using your actual policies, or an AI calling agent that books clinic appointments by following the same rules your receptionist uses. We test each skill against real examples, keep scripts to the minimum needed, and set up human approval steps wherever mistakes would be costly. Because the skills follow an open standard, you keep the freedom to change AI providers later.

The move of Agent Skills into a neutral foundation is a quiet announcement with a practical message: your business knowledge is worth capturing, and it can now be captured in a way that outlasts any single AI product. If you would like to explore which of your processes are ready for this, book a free 30-minute strategy call with Legacy Services at legacyservices.in/contact. We will look at your workflows and give you a clear, honest view of where an AI agent would help and where it would not.