mcp-builder
Pass
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches protocol documentation and SDK information from official sources including
modelcontextprotocol.ioand themodelcontextprotocolorganization on GitHub. These are trusted repositories for the technology provided by the skill. - [COMMAND_EXECUTION]: The
evaluation.pyscript is designed to run MCP servers as subprocesses using thestdiotransport to test their functionality. This execution is controlled by the user and is the primary purpose of the evaluation harness. - [PROMPT_INJECTION]: The evaluation harness processes tasks from XML files (
scripts/evaluation.py), creating a surface for indirect prompt injection. - Ingestion points: Task questions are read from user-provided XML files.
- Boundary markers: The
EVALUATION_PROMPTprovides instruction-based wrapping but no strict delimiters for the interpolated question. - Capability inventory: The evaluation agent has access to all tools provided by the MCP server under test.
- Sanitization: No sanitization is performed on the XML content.
- This is a low-risk finding associated with the tool's intended use for testing LLM tool-calling capabilities.
Audit Metadata