voice-of-customer-synthesizer

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly scrapes and ingests user-generated public content — e.g., Phase 1B "review-site-scraper for G2, Capterra, Trustpilot" and Phase 1C social listening ("site:reddit.com OR site:twitter.com") — and directs the agent to read and cluster those items into actionable recommendations, which could enable indirect prompt injection from untrusted third-party posts.