Check any skill, plugin, or package for malicious code before it runs on your machine. Get a clear RED / YELLOW / GREEN verdict in seconds.
$ repoguard scan https://github.com/sketchy/plugin
RepoGuard Report: plugin
Verdict: RED
Data Exfiltration █████████░ HIGH
Obfuscated Code ██████░░░░ MED
Install Scripts ████████░░ HIGH
Backdoors ░░░░░░░░░░ NONE
Privacy Violations ███░░░░░░░ LOW
Dependency Risks ░░░░░░░░░░ NONE
Filesystem Access █████░░░░░ MED
Supply Chain Red Flags ██░░░░░░░░ LOW
Top findings:
! postinstall script runs curl | bash (package.json:8)
! Reads .env and POSTs to Discord webhook (steal.js:12)
* Base64-encoded payload passed to eval() (utils.js:45)
Recommendation: DO NOT USE this repository. Skills and plugins run with your full permissions. Scan them before adding to your workflow.
MCP servers connect directly to your AI tools. Verify they're not exfiltrating your data.
Scan the source repo of any package before installing. Catch typosquats and malicious install scripts.
Extensions run inside your editor with access to your files. Check the source first.
Actions run in your CI/CD pipeline with access to secrets. Don't trust blindly.
Before you clone and run anything, let RepoGuard tell you what's inside.
npm install -g repoguardrepoguard scan https://github.com/author/skillGREEN = safe. YELLOW = review first. RED = don't install.
Use --json to get machine-readable output. Pipe it to Claude Code, ChatGPT, or any LLM for AI-powered false-positive filtering — no API key required.
$ repoguard scan https://github.com/author/mcp-server --json
# Full JSON output — feed it to any AI for deeper review