gemini-api

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs the agent to retrieve and use arbitrary web content (e.g., the Url Context tool and Search Grounding in references/structured_and_tools.md and the YouTube-from-uri example in references/text_and_multimodal.md), which causes the model to read untrusted, user-generated third‑party pages and use that content to shape responses.

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

• Potentially malicious external URL detected (high risk: 0.90). The MCP example invokes an external npm package at runtime via npx (args ["-y", "@philschmid/weather-mcp"]), which fetches and executes remote code and is used as a tool that can influence model prompts/behavior, so it is a runtime dependency that executes remote code.