mcp-builder
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes an evaluation harness (
scripts/evaluation.py) that can execute local shell commands to launch MCP servers during testing. This is a standard and expected feature for development tooling of this nature, and the execution is restricted to the local development environment under user initiation. - [DATA_EXPOSURE_AND_EXFILTRATION]: The documentation explicitly instructs users to store sensitive API keys in environment variables rather than hardcoding them, which aligns with security best practices for credential management.
- [EXTERNAL_DOWNLOADS]: The guide references official Model Context Protocol documentation and SDKs hosted on
modelcontextprotocol.ioandgithub.com/modelcontextprotocol. These are verified, official sources for the technology described and are considered safe. - [INDIRECT_PROMPT_INJECTION]: The skill includes a vulnerability surface evaluation (Phase 4), teaching developers how to test their servers against complex queries. It implements boundary markers (XML tags like
<response>,<summary>, and<feedback>) to help distinguish agent reasoning from tool outputs.
Audit Metadata