tech-leads-club/agent-skills

What it solves

It provides a secure, verified registry of "skills"—packaged instructions and resources—to extend the capabilities of AI coding agents. It addresses the security risks associated with open marketplaces, where many skills contain critical vulnerabilities, by offering a hardened, human-curated library.

How it works

Users can install skills via a CLI tool that fetches them from a CDN and installs them into the configuration of supported AI agents (such as Claude Code, Cursor, or Windsurf). The system uses a defense-in-depth approach including static analysis in CI/CD, content hashing, and Snyk Agent Scan to ensure safety. Additionally, it provides an MCP (Model Context Protocol) server that allows agents to search and load skills dynamically using progressive disclosure.

Who it’s for

Developers using AI coding agents who want to add specialized workflows, patterns, and knowledge to their assistants without compromising the security of their local environment.

Highlights

• Verified Security: Every skill is scanned with Snyk Agent Scan and undergoes human curation to prevent vulnerabilities.
• wget-like CLI: An interactive wizard and a set of command-line options for installing, updating, and removing skills across multiple agents.
• Broad Compatibility: Supports a wide range of agents including Tier 1 (Claude Code, Cursor, Copilot), Tier 2 (Aider, Antigravity), and Tier 3 (Enterprise tools like Amazon Q and Sourcegraph Cody).
• MCP Integration: An MCP server allows agents to browse and fetch skills on-demand via tools like list_skills and search_skills.