shopify-mcp

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The required Storefront MCP runtime path serves tool metadata at /.well-known/mcp.json and processes tool invocations via the POST handler /routes/api/mcp/tool-call.ts, where the outsider-provided tool input (e.g., search_catalog.input.query) is read and embedded into a GraphQL query and therefore ingested by the LLM-enabled agent workflow without selecting a specific pre-defined item first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The agent (OpenAI Operator / other AI agents) discovers and fetches the MCP configuration at "/.well-known/mcp.json" at runtime, and that fetched JSON directly defines the tools and behavior the agent will use, so the well‑known endpoint controls agent prompts and actions.