voice-of-customer

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow (Steps 2–4 in SKILL.md) explicitly uses the Perplexity tools (e.g., perplexity_ask) to fetch and quote user-generated content from online communities and review sites (Reddit, forums, social media, review sites) and then ingests those quotes into research-memory/customer-language.md for downstream use, so untrusted third‑party content can directly influence agent outputs and actions.