Supply and demand diverge on skills.sh: more than half of listings teach technical work, software engineering alone about a quarter, yet no category draws more than a fifth of installs. Software engineering leads at 18%, then AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → workflows at 15%.
Per listing, business operations skills draw 74% more installs than average, writing and documents 50% and cloud and infrastructure 42%. Software engineering skills draw about 30% fewer, education and productivity 39% fewer and research 55% fewer.
Installs are highly concentrated: nearly half of all skills were installed exactly once, 375 skills (0.04% of the registry) account for 62% of installs and the top 1.2% for 94%, yet no single agent skillA reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.Full definition → reaches 1% of installs.
Cross-industry skills, such as spreadsheet cleanup or website deployment, make up 66% of classified skills but 87.5% of installs, averaging 3.6 times the installs per listing of industry-specific skills.
Vercel cautions that install figures are aggregate registry counters, not unique people or independent choices, and that category results come from a classified sample. It expects skills to be judged by tests and benchmarks of effectiveness rather than installs.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
Why it matters
Quantifies how fast reusable agent instructions are being packaged and shared, giving engineers a benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → for how mainstream the skills pattern has become.
