Agent Skills Moves to the Agentic AI Foundation: What Changes
Vorec Team · 2026-10-23 · About 8 min read
On 22 October 2026, Anthropic donated Agent Skills, the open standard behind SKILL.md folders, to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation. The name, logo, GitHub organization and agentskills.io domain move to the foundation. Anthropic says you can keep using the standard, or contribute to it, without joining the AAIF.
If you write skills, nothing you have built stops working today. What changes is who decides where the format goes next, and the roadmap Anthropic has proposed for it: versioned releases, a formal proposal process, and security guidance for people who write and install skills. This article sets out what the announcement says, separates it from what it doesn't, and explains why it matters for a category that is easy to overlook: skills that teach agents to operate, record and document software.
All facts below come from Anthropic's announcement and the agentskills.io documentation, checked 23 October 2026. Where a point is our reading rather than the source's statement, we say so.
From Anthropic's 22 October announcement: more than 70 agent platforms have integrated Agent Skills, including ChatGPT, Cursor, GitHub Copilot and VS Code. Anthropic reports 1 million skills published and nearly 280 million installed. These are Anthropic's own figures; the post does not say how they were counted.
What exactly was donated
The announcement lists four things moving to the AAIF:
| Donated | What it is |
|---|---|
| The Agent Skills name and logo | The brand of the standard |
| The GitHub organization | Where the specification is developed (agentskills/agentskills) |
| The agentskills.io domain | Where the specification and documentation are published |
| The standard itself | The open format for SKILL.md folders |
Two points from the post are worth keeping straight:
- Participation does not require membership. Anthropic writes that you can continue to use the standard or contribute to it without joining the AAIF.
- Maintainers act as individuals. Engineers from Anthropic, Google, OpenAI and other companies will maintain the standard as individuals, with David Soria Parra and Mahesh Murag named as Lead Maintainers.
What the Agentic AI Foundation is
The AAIF is not new. Anthropic's December 2025 announcement describes it as a directed fund under the Linux Foundation, co-founded by Anthropic, Block and OpenAI, with support from Google, Microsoft, AWS, Cloudflare and Bloomberg. Its founding projects were the Model Context Protocol (MCP), Block's goose and OpenAI's AGENTS.md.
So Agent Skills now sits next to MCP in the same foundation. That pairing makes sense if you think of the two as complementary: MCP connects an agent to external tools and data, while a skill packages the procedural knowledge for a specific task. Our explainer on the Model Context Protocol covers the first half; What are Agent Skills? covers the second.
A quick refresher: what a skill is
Per the agentskills.io overview, a skill is a folder containing a SKILL.md file. That file holds metadata (a name and description at minimum) and instructions that tell an agent how to perform a specific task. A skill can also bundle scripts, reference material, templates and other resources in folders such as `scripts/`, `references/` and `assets/`.
Agents load skills through progressive disclosure. At startup they read only each skill's name and description, which is enough to know when a skill might apply, and load the full instructions when a task calls for them. Anthropic's announcement explains the origin: skills were developed while working on Claude Code, which handled general knowledge work but often lacked the procedural knowledge real business tasks need.
What the proposed roadmap covers
Anthropic's wording matters here. It says it has proposed starting with a roadmap covering three things. These are proposals for the project, not delivered features:
| Proposed item | What it would mean for you (our reading) |
|---|---|
| Versioned releases | A skill could state which version of the format it targets, and clients could say which versions they support |
| A formal proposal process | Changes to the format would go through a public, documented process instead of arriving ad hoc |
| Security guidance for people who write and install skills | Shared expectations for what a safe skill looks like and what to check before installing one |
The announcement adds that, as with MCP, the project will prioritize community input and transparent decision-making, and points to the project's Discord for feedback on proposed specification changes.
The security item is the one we'd watch most closely. A skill can bundle executable scripts, and with hundreds of millions of installs reported, installing a skill is closer to installing software than to reading a document. Anthropic's post doesn't say what the guidance will contain or when it will arrive.
What changes for skill authors today
Based on the announcement, very little changes in practice right now:
- Your existing SKILL.md files keep working. Nothing in the post deprecates or alters the current format.
- The specification lives where it did. agentskills.io and the GitHub organization are the same addresses; their owner is what changed.
- Contributing is open. You don't need to join the AAIF to propose changes or give feedback.
What we'd suggest, as our own reading rather than anything the announcement asks for:
- Keep skills close to the spec. If versioned releases arrive, skills that use only documented fields will be easiest to label and keep compatible.
- Write a clear description. Since agents decide whether to load a skill from its name and description, a vague description means the skill is loaded at the wrong time or not at all. The agentskills.io docs include a page on optimizing descriptions.
- Document what your scripts do. Whatever the security guidance ends up saying, a skill whose scripts are short and explained in the SKILL.md will be easier for others to trust.
Why this matters for skills that record and document software
Most coverage of skills focuses on coding and office documents. A growing group of skills does something else: it teaches an agent to operate an application and capture what happened, for a demo, a bug report or a tutorial.
That kind of skill carries procedural knowledge that is hard to fit in a prompt: how to plan a walkthrough, which actions to take in order, when to pause so a viewer can follow, and where the human reviews the result before anything is published. A portable format matters because the same skill can then work across several agent platforms, which is the "build once, use across any skills-compatible agent" argument the agentskills.io overview makes.
Two related trends are covered in more detail elsewhere on this blog: agents that operate a computer directly, and the SDK toolsets for browser and computer use that make those agents easier to build.
Where Vorec fits
Vorec ships a `record-tutorial` skill for Claude Code (and works with Codex through the same CLI). The agent researches the app and plans the walkthrough, drives the Vorec macOS recorder to capture it locally, and you review the take before anything is uploaded. Vorec then drafts narration matched to the captured workflow and generates the voiceover, so nothing is spoken into a microphone. In the editor you can add smooth cursor motion, cursor-follow zoom and click-based auto-zoom, and the same capture can also produce a written step-by-step guide. You can also record yourself with the Vorec app or upload an existing recording. Plans start at $9/month (Starter); a free trial is available (current allowances on the pricing page).
What we don't know yet
To keep this accurate as it ages, here is what the announcement does not say:
- Timing. No dates are given for versioned releases, the proposal process or the security guidance.
- Governance details. The post names Lead Maintainers but doesn't describe how decisions are made beyond the commitment to community input and transparency.
- How the adoption figures were measured. The 70+ platforms, 1 million skills and nearly 280 million installs are stated without a method.
We'll treat each of these as open until the project publishes them.
FAQ
Who owns Agent Skills now?
The Agentic AI Foundation, a directed fund under the Linux Foundation. Anthropic donated the name, logo, GitHub organization and agentskills.io domain on 22 October 2026.
Do I need to join the AAIF to use or contribute to Agent Skills?
No. Anthropic's announcement says you can continue to use the standard or contribute to it without joining.
Will my existing skills break?
Nothing in the announcement changes the current format. Future versioned releases are proposed, but no changes are announced.
Who maintains the Agent Skills standard?
Engineers from Anthropic, Google, OpenAI and other companies, acting as individuals, with David Soria Parra and Mahesh Murag as Lead Maintainers.
How is Agent Skills different from MCP?
MCP connects AI applications to external systems such as tools and data. A skill packages instructions and resources for a specific task. Both are now projects of the AAIF.
Which platforms support Agent Skills?
Anthropic says more than 70, including ChatGPT, Cursor, GitHub Copilot and VS Code.
Related reading: What are Agent Skills?, What is the Model Context Protocol?, Claude Code screen recording plugin and AI agent screen recording.
Sources: Anthropic, Donating Agent Skills to the Agentic AI Foundation (22 October 2026); Anthropic, Donating the Model Context Protocol and establishing the Agentic AI Foundation (9 December 2025); Agent Skills overview. All checked 23 October 2026.
Want the recording to narrate itself? Record with Vorec, or upload one you already have, and get a narrated tutorial plus a written guide. Start free. Trial includes up to 3 projects; exports carry a watermark.