mkt-icp-research

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md workflow (Phase 2: Multi-Platform Intelligence Gathering and Voice of Customer sections) plus Tools Integration and scripts/search-platforms.sh explicitly instruct the agent to fetch and scrape user-generated public content from Reddit, Twitter/X, Facebook groups, YouTube comments, G2/Capterra, Discord/Telegram and other open sites and to extract exact quotes that are then used to shape personas and messaging—so untrusted third‑party content is ingested and can directly influence actions.