mcp-builder

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The evaluation script scripts/evaluation.py launches local MCP servers as subprocesses using user-provided commands and arguments (e.g., python my_server.py) to facilitate integration testing.
  • [EXTERNAL_DOWNLOADS]: The documentation references official protocol specifications and SDK README files hosted on the modelcontextprotocol.io domain and its associated GitHub repository for research and implementation guidance.
  • [PROMPT_INJECTION]: The evaluation system processes external XML files containing test questions; these inputs are passed to the language model to test tool-calling capabilities, which introduces a surface for indirect prompt injection if the test files originate from untrusted sources. Evidence: Ingestion point in scripts/evaluation.py (XML parsing), lack of explicit boundary markers in the EVALUATION_PROMPT, and capability to call server tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:21 PM
Security Audit — agent-trust-hub — mcp-builder