mcp-developer
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: A deterministic detector flagged a potential injection pattern in
references/sharp_edges.md. Analysis confirms this is a false positive. The content is documenting how developers should handle malicious user input (e.g., "ignore previous instructions") to secure their own MCP servers, rather than attempting to manipulate the agent's behavior. - [DATA_EXFILTRATION]: No unauthorized data access or exfiltration patterns were detected. The skill explicitly instructs developers to avoid including secrets in tool outputs, demonstrating security-aware guidance.
- [REMOTE_CODE_EXECUTION]: The skill contains code templates for building MCP servers using established libraries. It does not perform any remote script downloads or unauthorized command executions. All examples use standard protocol transports like stdio and SSE.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies indirect prompt injection as a 'Sharp Edge' for developers and provides specific mitigation strategies, such as output sanitization and returning structured JSON instead of prose to isolate user-controlled data.
- [SAFE]: The skill is primarily educational, providing code snippets and architectural patterns for the Model Context Protocol. No malicious behaviors, obfuscation, or persistence mechanisms were found.
Audit Metadata