agent-scan

Security scanner for AI agents, MCP servers and agent skills.

What is agent-scan?

Security scanner for AI agents, MCP servers and agent skills.

Overview of agent-scan

What it is

agent-scan is a security scanner for AI agents, MCP servers and agent skills. It discovers installed agent components on a machine and scans them for prompt injections and vulnerabilities. The tool is developed by Snyk and is available as a Python package run via uvx or as a standalone binary. The CLI output is experimental and subject to change between releases. The tool does not publish an npm package.

Key features

  • Auto-discovers MCP configurations, agent tools, and skills on the machine.
  • Scans for prompt injection, tool poisoning, tool shadowing, toxic flows, malware payloads, untrusted content, credential handling, hardcoded secrets, suspicious downloads, malicious code, and secret detection.
  • Supports scanning of Claude, Cursor, GitHub Copilot, Windsurf, Gemini CLI, Amp, Amazon Q, and other agents.
  • Can scan a whole machine, a specific MCP configuration, a single agent skill, or all Claude skills.
  • Provides risk-based output (v0.6 and later) or issue-code output (v0.5.x).
  • Allows ignoring skill analysis with --no-skills.
  • Requires explicit user consent before contacting MCP servers during interactive runs.

Use cases

  • Security auditing of AI agent environments to identify prompt injection risks.
  • Scanning MCP server configurations for vulnerabilities before deployment.
  • Reviewing agent skills for malware payloads or credential leaks.
  • Enterprise security teams managing agent security risk at scale.
  • Evaluating untrusted or third-party MCP configurations in sandboxed environments.

Getting started

Before running a scan, sign up at Snyk and get an API token from https://app.snyk.io/account. Set the token as an environment variable:bashexport SNYK_TOKEN=your-api-token-hereRun with uvx (requires uv installed):bash# Scan the whole machineuvx snyk-agent-scan@latest# Scan a specific MCP configurationuvx snyk-agent-scan@latest ~/.vscode/mcp.json# Scan a single agent skilluvx snyk-agent-scan@latest ~/path/to/my/SKILL.md# Scan all Claude skillsuvx snyk-agent-scan@latest ~/.claude/skillsAlternatively, download a standalone binary from GitHub Releases.

License and pricing

The license is Apache-2.0. The tool is open source.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/snykagent-scan.svg)](https://yunodb.com/tools/snykagent-scan?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/snykagent-scan?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/snykagent-scan.svg" alt="Featured on YunoDB" height="20"></a>