content-research

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

• Third-party content exposure detected (high risk: 1.00). The skill's Layer‑1 agents (ad-intel-agent, listening-agent, competitive-content-agent, trend-agent) explicitly fetch and ingest public third‑party content — e.g., facebook.com/ads/library, ads.tiktok.com Creative Center, reddit.com, YouTube, G2/Capterra, and competitor websites via WebFetch/Exa/Tavily — and those findings are required inputs used to synthesize patterns and generate content briefs, so untrusted/user‑generated content can directly influence agent decisions and actions.