mcp-builder
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The evaluation harness (scripts/evaluation.py) and connection manager (scripts/connections.py) utilize the Model Context Protocol (MCP) SDK to launch and interact with local MCP servers. The script accepts shell commands and arguments via command-line flags to start these servers as subprocesses, which is the standard mechanism for testing local MCP integrations.
- [EXTERNAL_DOWNLOADS]: The documentation guides in SKILL.md instruct the agent to retrieve protocol specifications and SDK documentation from official sources, including modelcontextprotocol.io and the modelcontextprotocol GitHub organization. These resources are well-known, trusted entities within the AI development ecosystem.
- [PROMPT_INJECTION]: The evaluation loop in scripts/evaluation.py ingests data from the tools of the MCP server being tested and provides it to the LLM agent. This constitutes an indirect prompt injection surface where a malicious server could attempt to influence the agent's summary or feedback. However, this is a known risk inherent in testing tool-use capabilities, and the script is intended for use in controlled development environments.
Audit Metadata