Skip to content

Security audit findings — tool description injection + missing output sanitization #1682

@manja316

Description

@manja316

Hi MCPJam team,

I recently ran a security audit on MCP Inspector as part of research on MCP tooling security posture.

Found a couple of items worth flagging:

1. Tool description injection risk
The inspector renders tool descriptions from any connected MCP server without validating them against adversarial patterns. As a debugging/development tool, developers often connect to untrusted or third-party servers — a poisoned tool description could redirect the inspector's LLM into leaking session data or executing unintended operations.

2. Missing output sanitization
Tool outputs displayed and forwarded to the model context aren't scanned for injection patterns. Since the inspector is commonly used to test server behaviour, this is a high-value target for demonstrating the attack class.

Both are in a full audit report — 8-page PDF with CVSS ratings, EU AI Act mapping, and remediation steps — for $29 at luciferforge.github.io/mcp-security-audit/

Demo: https://luciferforge.github.io/mcp-audit-reports/

— Lucifer / LuciferForge Security

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions