dark-funnel

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

• Third-party content exposure detected (high risk: 0.85). The skill’s workflow ingests “dark funnel” inputs such as Reddit/community mention frequency, AI chatbot citation data, podcast attribution metrics, and branded search volume trends; these are outsider-authored free text (e.g., Reddit/forum posts, chatbot-retrieved snippets, podcast/community discussion text) that can be fetched/aggregated at runtime and then fed into the agent’s LLM context for analysis.