mcp-gateway

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The StdioTransport implementation utilizes tokio::process::Command to spawn and manage local subprocesses. This is used in the configuration example to run the filesystem server.
  • [EXTERNAL_DOWNLOADS]: The configuration demonstrates fetching and executing the @modelcontextprotocol/server-filesystem package via npx -y. While targeting a well-known organization, this pattern involves downloading and running code at runtime.
  • [DATA_EXFILTRATION]: The HttpTransport, SseTransport, and WebSocketTransport implementations enable communication with remote URLs. These transports are designed to transmit authentication credentials (Bearer tokens and API keys) to the configured MCP server endpoints.
  • [PROMPT_INJECTION]: The gateway's architecture for processing tool outputs from remote servers creates a surface for indirect prompt injection.
  • Ingestion points: The skill reads and parses JSON-RPC responses from external HTTP, SSE, WebSocket, and Stdio sources.
  • Boundary markers: No specific delimiters or safety instructions are shown to isolate tool output from the agent's control flow.
  • Capability inventory: The gateway has the ability to spawn processes (StdioTransport) and perform network requests (multiple transports).
  • Sanitization: The implementation focuses on transport logic and does not show explicit sanitization or validation of the content returned by external tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 09:48 AM
Security Audit — agent-trust-hub — mcp-gateway