GEO

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

• Third-party content exposure detected (high risk: 0.80). The audit workflow (audit.md), especially the "Phase 2: Multi-Model Testing" and "Automation Tips" / Example agent task, explicitly instructs running queries against external AI models and ingesting their responses (and the strategies.md recommends tracking content on public sources like Reddit, Hacker News, and Wikipedia), so the agent is expected to consume untrusted, user-generated third‑party content and use it to drive analysis and follow-up actions.