world-cup-predictor

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

• Third-party content exposure detected (high risk: 0.85). The runtime workflow ingests outsider-authored free text via scripts/collect_signals.py (news/weather and Polymarket/odds manual-check targets), which then gets embedded into the agent’s LLM context for subagents and synthesis; this includes arbitrary public-page content and scraped text from non-user sources.