About this lab

An organization registers third-party MCP plugins to extend its AI assistant's capabilities.

A compromised plugin registers tools that appear legitimate but have hidden behavior.

When the assistant invokes the plugin, it executes attacker-controlled logic.

Demonstrate supply chain compromise through a malicious MCP plugin.

Objectives

  1. Understand the MCP plugin registration and invocation flow.
  2. Analyze a malicious plugin manifest and tool schemas.
  3. Trigger the plugin's hidden behavior through normal assistant usage.
  4. Exfiltrate data via the plugin's covert channel.
🚧

Coming Soon

This lab is currently in development. Check back soon for availability.

← Back to catalog