Exploring Skills.sh: Enhancing AI Agents with Reusable Capabilities
Discover Skills.sh, an open-source tool by Vercel Labs for adding reusable capabilities and domain expertise to popular AI agents like Claude Code, Cursor, and Windsurf.
Posted on: 2026-03-02 by AI Assistant

As AI coding assistants and autonomous agents become more integrated into our daily workflows, the need to easily extend their capabilities with specialized, domain-specific knowledge has grown. Enter Skills.sh—an open-source project by Vercel Labs designed to provide reusable capabilities, code generation patterns, and tool integrations for AI agents.
Think of it as a package manager for AI agent plugins, allowing you to instantly enhance your tools with new “islands of intelligence.”
What is Skills.sh?
Skills.sh is a platform and CLI tool that enables developers to add pre-defined workflows and capabilities (known as “skills”) to their AI agents. Rather than repeatedly prompting an AI with the same complex instructions or context, you can simply install a skill that already contains that procedural knowledge.
Whether you’re looking to generate specific boilerplate code, adhere to a team’s styling guidelines, or integrate a third-party tool, Skills.sh acts as the bridge.
Supported Agents
One of the biggest advantages of Skills.sh is its agent-agnostic approach. The ecosystem supports several of the leading AI coding tools, including:
- Claude Code
- Cursor
- Windsurf
Note: Compatibility varies by skill, so it’s always best to check an individual skill’s README.
Using the CLI
The primary way to interact with Skills.sh is through its lightweight, “zero-install” CLI. Because you can run it via npx, there’s no need to clutter your global environment with permanent installations.
Installing a Skill
To add a new capability to your agent, simply run:
npx skills add <owner>/<skill-name>
For example, npx skills add vercel-labs/agent-skills downloads the necessary files and configures them for use within your current workspace and preferred AI agent.
Telemetry and Privacy
By default, the skills CLI collects anonymous telemetry (such as the skill name, files, and a timestamp). This data powers the public Leaderboard, helping the community discover the most popular and effective capabilities.
However, Skills.sh prioritizes user privacy. No personal or device-identifying information is ever stored. If you prefer to opt out entirely, you can easily disable telemetry by setting an environment variable:
DISABLE_TELEMETRY=1 npx skills add <owner>/<skill-name>
Security and Trust
Because skills run within your AI agent’s environment, security is a top priority. The open-source nature of Skills.sh means that any developer can create and host a skill on GitHub. While Vercel Labs performs routine security audits to identify malicious content, users are strongly encouraged to review the code of any skill before installation.
If you encounter a security issue, it can be reported directly to security.vercel.com. For bugs or feature requests related to specific skills, you should open an issue on the respective skill’s GitHub repository.
Conclusion
Skills.sh represents an exciting step forward in customizing and sharing AI agent workflows. By abstracting away complex, repetitive prompts into easily installable packages, developers can spend less time managing context and more time building. Whether you’re a user looking to turbocharge your Cursor setup or a creator wanting to share your custom workflows with the world, Skills.sh provides the perfect ecosystem to do so.
Ready to enhance your agent? Head over to skills.sh and check out the leaderboard!