harper-mcp
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes building an interface that processes external input from MCP clients. Ingestion points include arguments in mcpTools methods, mcpPrompts render functions, and URI template parameters in mcpResources. Boundary markers are addressed by instructions to use context.user for manual authorization. The capability inventory includes access to internal database tables. Sanitization is covered by specific warnings against path traversal in URI templates.
- [DYNAMIC_EXECUTION]: The skill explains how to register custom server-side logic for runtime invocation via mcpTools methods and mcpPrompts render functions.
- [SAFE]: All findings related to data ingestion and dynamic execution are descriptive of the Harper MCP architecture being documented and do not constitute malicious code. The skill provides proactive security guidance and hardening checklists for production deployments.
Audit Metadata