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's Phase 1 workflow explicitly instructs scraping public, user-generated review sites (G2, Capterra, Trustpilot) and social platforms (Twitter, Reddit) using tools like review-scraper/twitter-scraper/reddit-scraper, which the agent ingests and uses to drive theme clustering and recommendations, exposing it to untrusted third-party content that could carry indirect prompt injections.