mcp-builder
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to fetch the latest protocol specifications and SDK README files from modelcontextprotocol.io and the official Model Context Protocol GitHub repository. These are trusted, well-known sources for the technology being documented.
- [COMMAND_EXECUTION]: The provided testing script (
scripts/evaluation.py) is designed to execute a command-line string provided by the user to launch a local MCP server for evaluation. This is a functional requirement of the tool's purpose as a development testing harness. - [INDIRECT_PROMPT_INJECTION]: The skill's evaluation harness processes external data that could theoretically contain instructions.
- Ingestion points: Evaluation questions are read from user-provided XML files, and tool outputs are retrieved from the MCP server being tested.
- Boundary markers: The system prompt in
evaluation.pydefines specific XML tags (<summary>,<feedback>,<response>) to structure the agent's output and minimize unintended behavior. - Capability inventory: The script can execute local commands (to start the server) and call the tools defined by the server being evaluated.
- Sanitization: Data from tools and files is stringified before being passed to the assistant.
Audit Metadata