Labsco
MCP SERVER

MCP Prompt Injection Scanner

by CSOAI-ORG

Scan for prompt-injection patterns using 30+ canonical detection rules across five severity tiers.

Agent Guardrails, Content Safety & Audit
Summary
Severity tiers are what make a rule set usable.

Thirty rules that all shout equally loudly produce a report nobody reads; sorting them into five tiers is the difference between a scanner and a noise source. The repository documents the rule set and the threat class it targets, and leaves the tool surface to be discovered from the running server.

What it is

A Python server that scans for MCP injection. It ships 30+ canonical detection rules sorted into 5 severity tiers, aimed at the MCP remote-code-execution class described in April 2026 — so findings arrive ranked rather than as a flat list you have to triage.

What you get

Rule-based scanning with severity already attached: 30+ canonical rules across 5 tiers, so a high-severity hit is distinguishable from a stylistic one without reading every result.

Requirements

`pip install meok_mcp_injection_scan_mcp`, or add the PyPI package `meok-mcp-injection-scan-mcp` to your client over stdio (registry version 1.3.2). MIT-licensed, and installable through Smithery with `npx -y @smithery/cli@latest install meok-mcp-injection-scan-mcp --client claude`.

Setup effort

One command — pip install meok_mcp_injection_scan_mcp