The Most-Installed Skill Isn't a Coding Skill. That's the Whole Story.

When I first saw that find-skills from vercel-labs had 1.3 million installs — more than any React skill, more than any deployment skill, more than any developer productivity skill — I assumed it was a distribution quirk.

It's not. It's a signal.

What find-skills Does

The find-skills skill does exactly one thing: given a problem or intent, it searches the skills.sh registry and returns relevant skills that could help.

It's a recommendation engine, but at the level of agent capabilities rather than content. You describe what you're trying to do. It tells you which skills exist to help you do it.

That's it. There's no code generation. No workflow automation. No specialist domain knowledge. Just: here are the tools that exist, here is what they do, here is how to install them.

The Interpretation Nobody Is Saying Out Loud

A meta-tool being the most-installed tool means one thing: the primary problem people are solving right now isn't any specific task. It's figuring out what's possible.

We're not in the productivity phase of AI agent tooling. We're in the discovery phase.

That distinction matters more than it sounds. In a mature tooling ecosystem, the most-used tools are the ones that do the most useful work — compilers, test runners, deployment tools. The meta-tools — package registries, dependency managers, documentation systems — are infrastructure. They're invisible. Nobody installs a package manager because it's their favorite tool; they install it to get to the tools they actually need.

The fact that find-skills has 1.3M installs while the downstream skills it recommends have install counts in the tens and hundreds of thousands means the funnel hasn't narrowed yet. People are searching. They haven't fully landed.

What Discovery-Phase Adoption Looks Like

In the early days of npm, a significant portion of activity was npm search and npm info — people trying to understand what existed before they knew what to install. The install numbers came later, once the use cases became clear and the patterns solidified.

Skills are in that phase right now. The ecosystem is large enough to be useful but young enough that most people don't have a mental model of what's in it. The natural response to that situation is to install the tool that helps you build the mental model.

find-skills is that tool. Its install count is a measure of orientation activity, not deployment activity.

The More Interesting Question

What happens to find-skills' relative position in six months?

If the ecosystem matures — if skills for specific domains become well-known, if patterns emerge that tell builders exactly which skills they need for which situations — the meta-skill gets less necessary. The number of new people entering the ecosystem keeps growing, but each person needs it for a shorter time.

If the ecosystem keeps expanding faster than the mental models do, find-skills stays on top. More things to discover means more ongoing demand for the discovery tool.

I think the latter is more likely, at least for now. The rate at which new skill collections are being published — React skills, database skills, design engineering skills, domain-specific marketing skills — is outpacing the rate at which any individual developer can track the landscape.

That means find-skills isn't a transitional tool. It's infrastructure. The meta-layer that makes the rest of the ecosystem navigable.

1.3 million installs for a tool that just finds other tools. That's the clearest signal in the whole registry about where we are.

---

Part of the AI Skill Daily series — skills worth understanding, one at a time.