research-note

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

• Third-party content exposure detected (high risk: 0.75). The skill performs runtime web searches in Phase H (“News & Catalysts via WebSearch”) and then ingests the resulting outsider-authored web page/snippet text into the LLM context via news_timeline, forward_catalysts, and policy_backdrop (indirect prompt injection risk from public web content).