mcp-builder
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (LOW): The skill directs the agent to fetch documentation and README files from external sources such as
modelcontextprotocol.ioand GitHub repositories (modelcontextprotocol/python-sdk,modelcontextprotocol/typescript-sdk). While these are the official repositories for the protocol, the GitHub organization is not on the predefined trusted list. - PROMPT_INJECTION (LOW): Indirect Prompt Injection Surface (Category 8). The skill instructs the agent to ingest and process external Markdown documentation without explicit boundary markers or instructions to disregard embedded instructions.
- Ingestion points:
modelcontextprotocol.io/sitemap.xml, GitHub README files, and various protocol specification pages. - Boundary markers: Absent; the skill does not specify delimiters for fetched content.
- Capability inventory: The skill context encourages the use of
WebFetchand command-line tools for building and testing servers. - Sanitization: Absent; fetched documentation is processed directly as guidance for the LLM.
- COMMAND_EXECUTION (LOW): The guide recommends executing local commands such as
npm run build,npx @modelcontextprotocol/inspector, andpython -m py_compile. While standard for development, these constitute a command execution surface if the environment is misconfigured or if the build scripts are compromised.
Audit Metadata