adk-eval-guide

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md (see the "Deep Dive: ADK Docs" and related reference sections) explicitly instructs using WebFetch to fetch public web pages (e.g., https://google.github.io/adk-docs/...), so the agent is expected to ingest external public web content that could materially influence evaluation decisions.