Agent Security Scanner

Scans AI agents for tool-calling vulnerabilities such as prompt leaks, hijacking, and injections.

What you can do

Agent Security Scanner is a local MCP server that lets an assistant run security scans against AI agents, focusing on tool-calling weaknesses. According to its description, it checks for prompt leaks, hijacking, injections, and similar issues. No individual tools are listed in the manifest, so the exact commands and outputs are not documented here.

When to use it

Consider it when you are building or testing an AI agent that calls tools and want a quick way to probe for prompt leakage, hijacking, or injection weaknesses. It is aimed at security review of agent behavior rather than general application scanning.

Requirements

Runs as a local process on the user's machine, installed from the PyPI package agentvuln over stdio. It is desktop only and has no web executable. No authentication, environment variables, or headers are declared, so no credentials appear to be needed.