MCPSafe deep scan: 3 high-severity findings, score 79/100 — badge available

[mcpsafe-gh]

Hi team 👋

We ran a free MCPSafe deep scan on mcp-adapter using a 5-LLM consensus engine and wanted to share the results directly.

Deep Scan Results (AIVSS Score)

Severity	Count
🟠 High	3
🟡 Medium	5
Score	79 / 100 (Grade C)

📋 Full report with evidence, finding details & PDF export:
https://mcpsafe.io/registry/github/vercel/mcp-adapter

MCPSafe detects MCP-specific threats that traditional SAST misses: prompt injection via tool descriptions, tool poisoning, secret exfiltration paths, overbroad schemas — scored with AIVSS (AI Vulnerability Severity Score). Uses 5-LLM consensus so only multi-model-confirmed findings are flagged.

Add the badge to your README:

[![MCPSafe](https://api.mcpsafe.io/badge/github/vercel/mcp-adapter.svg)](https://mcpsafe.io/registry/github/vercel/mcp-adapter)

The badge auto-updates with each scan. Free, no signup required for public repos. Happy to answer questions about the findings.