4,200 skills claim to do task X. Which one works?
By 2026, the SKILL.md ecosystem had exploded — 4,200 public skills on claudemarketplaces.com, 1.2M on skillsmp.com, plus Anthropic's own enterprise marketplace. None of them answered the only question a developer adopting Claude Code actually asks.
“Of the 47 skills that claim to do task X, which one actually works?”
Existing directories rank by stars and installs — popularity, not quality. Versuz runs every skill against the same task suite, has three frontier judges grade the outputs, and publishes a ranking based on actual performance.
Versuz in your terminal.
Two surfaces beyond the web app : a CLI for direct installs, and an MCP server so Claude Code itself can browse the registry inline. Both in public beta.
npx versuz
Interactive prompt-driven CLI. Search, browse, inspect, install. Free items download directly to .claude/skills/<slug>/SKILL.md ou ./CLAUDE.md. Premium items return a buy URL — purchase first, then install.
npx versuz # interactive npx versuz search pdf npx versuz info <slug> npx versuz install <slug>Source on GitHub ↗
@versuz/mcp
An MCP server that hooks Versuz into Claude Code as native tools. Ask Claude "find me a SQL migration skill and install it" — it searches, inspects, downloads, all inline. 5 tools : search, list_skills, list_claude_md, get, install.
claude mcp add versuz npx -y @versuz/mcp
# .mcp.json
{
"mcpServers": {
"versuz": { "command": "npx", "args": ["-y", "@versuz/mcp"] }
}
}Source on GitHub ↗What's next.
Document benchmark.
PDF/document extraction skills, 30 tasks, 3 judges, daily Bayesian Elo. Read-only public leaderboard.
Multi-vertical + monetisation.
More categories (SQL, data, web scraping). Authors can monetise — 70% revenue share when companies install via Versuz.
Real-time battles.
User-submitted match-ups, live judging, voting. Chatbot-Arena-style for agent skills.