mcp-use-code-review

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill consists entirely of markdown files providing documentation, best practices, and code examples. There are no executable scripts or active components that perform operations on the host system.
  • [NO_CODE]: The skill provides no functional code or tool definitions; it is purely informational, designed to educate the agent on the architecture and usage of the mcp-use library.
  • [PROMPT_INJECTION]: The instructions were analyzed for malicious patterns, such as bypass commands or role-play jailbreaks. No such content was found. The guidance provided (e.g., '6 AI Derailment Patterns') is focused on improving the quality and safety of the agent's code reviews.
  • [DATA_EXFILTRATION]: No sensitive data access or exfiltration patterns were detected. Documentation for authentication and session management correctly uses placeholders and best practices for environment variable management.
  • [REMOTE_CODE_EXECUTION]: The documentation describes standard MCP functionality where servers can run subprocesses or connect via HTTP/WebSocket. These are explained as legitimate configuration options for the SDK and do not constitute an instruction for the agent to execute unauthorized code.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 07:19 PM