Agent Prompt Injection Firewall MCP Server

This prompt injection firewall MCP has a reasonable security foundation with proper pattern-based detection and tiered access control, but contains several concerning issues. The code attempts to load a shared auth engine from a hardcoded local path that will fail gracefully, implements a fallback auth mechanism, and includes network calls to an attestation API. The main vulnerabilities are: (1) unauthenticated access to core scanning functionality when auth engine is unavailable, (2) arbitrary regex execution via user-supplied patterns without proper sandboxing, (3) network calls sending user data and API keys to external endpoints, and (4) subtle logic flaws in the free-tier rate limiting that could be bypassed. Supply chain analysis found 3 known vulnerabilities in dependencies (0 critical, 3 high severity). Package verification found 1 issue.

4 files analyzed · 14 issues found

Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.