1M Agent Skills in 7 Months: What skills.sh Means for Swiss SMBs
In short
An agent skill is a reusable, plain-language instruction that tells an AI agent how to complete a task. skills.sh reached 1 million skills in 7 months – faster than any earlier platform. For Swiss SMBs, that means: start documenting proprietary process knowledge as skills now to make your agents effective tomorrow.
The milestone: 1 million agent skills in 7 months
Anthropic introduced agent skills in October 2025, and Vercel launched skills.sh in January 2026. Seven months later, the count stands at one million – faster than GitHub took to reach one million repositories (27 months), the Apple App Store (over five years) and npm (over nine years). With close to 280 million installations, agent skills are no longer an experiment; they are a new layer of enterprise IT.
1M
Agent skills on skills.sh within 7 months
What an agent skill actually is
An agent skill is a reusable set of instructions written in plain language, typically as a SKILL.md file with YAML frontmatter. It tells an AI agent how to complete a specific task – for example reviewing a quote, preparing a supplier evaluation or classifying a service case.
- Skills separate generic agent capabilities from company-specific judgment.
- They are not code, but documented ways of working.
- They can be versioned, tested and shared across teams.
The first million skills taught agents what everyone knows. The next million will teach your agents what only you know.
What the distribution tells Swiss SMBs
87.5 percent of installations are for cross-industry skills. Supply is still heavily technical – software engineering accounts for 25 percent of listings but only 18 percent of installations. Demand is broader: agent workflows 15 percent, business operations 11 percent, writing 11 percent. Business operations skills earn 74 percent more installs per listing than average.
The bottleneck is documentation, not technology
Company knowledge lives in people's heads, emails and unwritten rules. Agent skills force you to make that knowledge explicit – that is where the real value for your company lies.
The strategic shift: from generic to proprietary knowledge
Concentration is extreme: 0.04 percent of skills – 375 out of one million – account for 62 percent of all installations. That does not mean the rest are worthless. It means generic capabilities are commoditizing fast, while company-specific knowledge becomes the differentiator.
Tests and benchmarks are coming
The next generation of skills will be measurable. Instead of just describing how to do a task, skills will soon include tests and benchmarks that show whether the agent performs better with the skill than without. That means you can prove quality, not just claim it.
What Swiss SMBs can do now
The starting point is smaller than many expect. Begin with a single, well-defined process your team currently handles repetitively and with errors. Describe it in plain language precisely enough that a new employee could execute it. As the Swiss podcast on AI in mid-sized companies argues, AI adoption is a leadership topic, not merely an IT project.
- Create a process inventory: Which three to five tasks consume the most time?
- Choose a process with clear inputs and outputs.
- Document the workflow in plain language, including decision rules.
- Test the skill on a small task and iterate.
The first step is smaller than you think
You do not need an in-house AI team. What matters more is data access: before an agent can take over a process, it must be able to reach the right systems.
Conclusion: secure your process knowledge now
The first million skills taught agents what everyone knows. The next million will be built from knowledge that only your company has. Swiss SMBs that start documenting their processes as skills today create a lead that will be hard to close later.
Frequently asked questions
- What exactly is an agent skill?
- A reusable instruction in plain language, typically as a SKILL.md file with YAML frontmatter. It describes how an AI agent should complete a specific task, without being program code.
- Why is 1 million skills in 7 months significant?
- Because no earlier developer platform reached that mark so quickly. GitHub took 27 months, the Apple App Store over five years, npm over nine years. The pace signals that agent skills are becoming a new standard layer.
- Are agent skills suitable for non-technical people?
- Yes. Skills are written in plain language, not code. What matters is describing process knowledge clearly – something subject-matter experts often do better than developers.
- How does a Swiss SMB start with agent skills?
- Start with a single, well-defined process. Describe the workflow precisely enough that a new employee could execute it, then test the skill on a small task.
- What changes with tests and benchmarks for skills?
- It will become measurable whether an agent performs a task better with a skill than without. Companies can prove quality instead of just claiming it.
Sources
Would you like to explore this topic for your company?
Check Availability