gemini-api

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

• Third-party content exposure detected (high risk: 0.85). The skill’s REQUIRED workflow includes “Always fetch fresh docs before generating or advising on code” and explicitly instructs fetching Google documentation URLs at runtime; these are outsider-authored public web pages (Google docs) that can be ingested as readable text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to "Always fetch fresh docs" at runtime and to "Fetch the model overview" (e.g. https://ai.google.dev/gemini-api/docs/models), so content fetched from that URL will be used at runtime to determine model IDs/capabilities and thus directly influence prompts/configuration.