2 million installs in under a month. That is the adoption rate Vercel is reporting for Skills, a new tool that gives AI agents a standardised way to do things.
Hello and welcome back to AI SIX. Six minutes, one topic, and everything you need to stay ahead on AI. I’m on a mission to upskill half a million people, and today’s episode is a good example of why that mission matters. Because what Vercel has just built tells us something important about where Agentic AI is actually heading.
So what is Skills, and why does it matter?
Vercel is a web development platform used by millions of developers. In January, they launched Skills, a command-line tool that lets developers install pre-built capabilities for AI agents the same way they would install any piece of software. You type a simple command, and your agent gets a new ability.
Alongside that, they launched skills.sh, a public directory and leaderboard where developers can browse, discover, and track which agent capabilities are being adopted. Think of it as an app store, but for what AI agents can do.
The idea actually started with one engineer, Shu Ding, sitting down on a weekend and writing up everything he knew about React and web development. The challenge was how to get that knowledge into AI agents quickly, without waiting for it to show up in the next model training cycle, which could take years. Skills became the answer. A simple, fast way to give agents up-to-date, specialist context on demand.
Vercel’s CTO Malte Ubl, described it simply: it is a package manager for agent context.
What Skills actually change
Here is the key distinction. Most AI agents generate instructions on the fly, working things out as they go. That means unpredictable outputs, especially when the agent hits a gap in its training data.
Skills separate reasoning from execution. Instead of letting an agent invent its own approach every time, you give it a controlled set of predefined commands. The agent does not improvise. It draws on verified, tested capabilities that developers have already built and shared.
What can a skill actually do? It can read or modify files. Run build steps. Interact with APIs. Query project data. All in a predictable, auditable way. The agent gets a toolbox instead of having to forge tools from scratch each time.
As of the Skills Night event in San Francisco on February 17th, the ecosystem had hit 2 million CLI installs and over 69,000 unique skills tracked in the directory. Skills already work across 18 AI agents, including Claude Code, GitHub Copilot, Cursor, and Windsurf. Those are the tools developers are using daily right now to build AI-powered products and services.
The trust problem nobody wants to say out loud
Here is where I want to slow down, because the speed of adoption is raising a real governance question.
Vercel provides verified official skills. But the ecosystem is open. Any developer can publish a skill. And at the Skills Night event, a security researcher from Socket demonstrated exactly why that matters. He showed a skill that looked completely clean on the surface, but buried inside was a file that opened a remote connection on install. You would never catch it without examining every file manually.
That is not a theoretical risk. That is a live example from launch week.
Vercel has moved quickly. They have announced security partnerships with Gen, Socket, and Snyk to audit every skill that comes into the ecosystem. They are building an audits leaderboard so organisations can see the security assessment for any skill before they install it.
But here is what organisations need to understand. Skills separate reasoning from execution, which improves predictability. Predictability is not the same as safety. If an agent installs a skill package that touches sensitive APIs, modifies customer data, or bypasses your authentication controls, who is accountable? Your organisation needs an answer to that question before you deploy, not after.
My Take
2 million installs in weeks tells you the market was waiting for this. Developers were already building agent workflows. They just lacked a shared standard for distributing capabilities safely. Skills filled that gap, and fast.
But here is what I keep coming back to. Value in AI is shifting. For two years, the conversation has been about which model is biggest, which benchmark is highest. Vercel did not build a model. They built the infrastructure that makes models useful in practice, the layer that turns AI capability into AI action. Meta understood this when they bought Manus. Vercel understands it too.
For anyone building or buying AI right now, the lesson is the same. The model matters. But the ability to reliably execute multi-step tasks in the real world, with auditable, verified capabilities, might matter more.
The Agentic AI era is not coming. It is operational. The organisations winning are the ones building on shared, auditable infrastructure, and governing it with the same rigour they bring to any other enterprise system.
If you want to build your AI literacy and skills, I can help
📚 Grab AI Playbook on Amazon, the perfect podcast companion.
📰 Subscribe to Human in the Loop, my LinkedIn Newsletter
🎓 Take the CPD Accredited AI Playbook Diploma
📞 Book a call at digitaltraining.ie
Sources
Vercel Changelog — Introducing Skills (Jan 20, 2026): https://vercel.com/changelog/introducing-skills-the-open-agent-skills-ecosystem
Vercel Blog — Skills Night: 69,000+ ways agents are getting smarter (Feb 20, 2026): https://vercel.com/blog/skills-night-69000-ways-agents-are-getting-smarter
InfoQ — Vercel Introduces Skills.sh (Feb 4, 2026): https://www.infoq.com/news/2026/02/vercel-agent-skills/
skills.sh directory and leaderboard: https://skills.sh
Key Statistics
Skills launched: January 20, 2026
CLI installs as of Feb 17, 2026: 2 million
Unique skills tracked: 69,000+
AI agents supported: 18+, including Claude Code, GitHub Copilot, Cursor, Windsurf
Security partners: Gen, Socket, Snyk
Skills Night event: San Francisco, February 17, 2026
About the Podcast Host
AI and Communications Specialist | Author | Podcaster | International Speaker
Joanne Sweeney is a recognised leader in digital marketing and digital transformation, specialising in the practical application of AI to enhance communication strategies. The founder of Digital Training Institute, Joanne, empowers leaders and communications professionals to leverage AI responsibly and effectively, ensuring transparency, trust, and tangible results.
She completed the Harvard Kennedy School Executive Programme in AI and Public Policy this summer before launching her fourth book AI Playbook, in September. Joanne has also published two editions of Public Sector Marketing Pro, which were preceded by her first title, Social Media Under Investigation Law Enforcement and the Social Web.
Her books have become go-to resources for governments, public sector and policing agencies, as well as marketing and communications professionals, all navigating the challenges of the changing Internet.
Joanne is the host of the AI SIX Podcast, a daily six-minute podcast that provides actionable insights, expert interviews, and practical advice to help professionals navigate the new world of work.
