vercel-labs/skills

It solves the fragmentation of AI agent customization by providing a single, standardized way to distribute and install skills across many different coding agents. It eliminates the need to manually copy configuration files or prompts into each agent's specific directory, and it supports private repositories and multiple source formats, making skill sharing and reuse straightforward.

Developers can build and share reusable skill packages that encode best practices, design guidelines, or domain-specific instructions, then install them into any supported agent with a single command. Real-world use cases include enforcing coding standards, providing frontend design guidelines, or adding specialized workflows (e.g., Convex best practices) to agents like Claude Code, Codex, or Cursor. The ceiling is high: with support for 75+ agents and flexible source resolution, it can become the standard package manager for AI agent skills, enabling a rich ecosystem of community-contributed skills.