mcp-server-patterns

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references the installation of the official @modelcontextprotocol/sdk and the zod validation library via npm. These are standard, well-known dependencies for the intended purpose.
  • [COMMAND_EXECUTION]: The documentation describes how to implement tools that may execute commands or search functionality, which is a core feature of the Model Context Protocol. It explicitly recommends using Zod for strict input schema validation to ensure safety.
  • [PROMPT_INJECTION]: The skill addresses the ingestion of external data by providing a framework for building MCP tools and resources. It implements security best practices by documenting the following: 1. Ingestion points: Data enters the system through tool input parameters defined in inputSchema and resource identifiers (URIs). 2. Boundary markers: The skill emphasizes defining strict input schemas for every tool to delimit expected data. 3. Capability inventory: Describes the registration of tools (registerTool) and resources (registerResource) across the SDK implementation. 4. Sanitization: Recommends the use of Zod for input validation and the return of structured error messages instead of raw stack traces to prevent data leakage.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 10:24 PM
Security Audit — agent-trust-hub — mcp-server-patterns