build-clean-mcp-architecture
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a defensive toolset designed to improve the security and structure of AI agent extensions (MCP servers). It enforces strict validation boundaries using Zod, centralized configuration to prevent secret leakage via environment variables, and standardized error handling to prevent internal data exposure.
- [DATA_EXPOSURE_DEFENSE]: The skill explicitly instructs on how to prevent data exfiltration and exposure through mandatory redaction of DSNs, provider names, and signed URLs in the
McpPresenterand gateway layers (seereferences/presenter-and-tool-response.mdandreferences/gateways-and-ports.md). - [DYNAMIC_CONTEXT_INJECTION]: The skill uses the
!syntax inSKILL.mdto run local audit scripts (e.g.,scripts/audit-layer-imports.sh). These scripts are bundled with the skill and perform static analysis (usinggreporfind) on the user's project to ensure architectural compliance. They do not perform network operations or access sensitive system files. - [REMOTE_CODE_EXECUTION]: No remote code execution patterns were found. All scripts provided are local and intended for development-time auditing.
Audit Metadata