gemini-live-api-dev

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

• Third-party content exposure detected (high risk: 0.75). The required workflow includes a “Documentation Lookup” fallback that at runtime fetches public web content from URLs (e.g., https://ai.google.dev/.../*.md.txt) and/or uses an MCP tool’s returned documentation text, which can contain free-form outsider-authored content that may be inserted into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's "Documentation Lookup" section explicitly instructs fetching the docs index and then specific .md.txt pages at runtime (e.g., https://ai.google.dev/gemini-api/docs/llms.txt), which would inject external documentation into the agent's context and thus control its prompts.