keyword-serp

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

• Third-party content exposure detected (high risk: 0.85). The required workflow calls SERP endpoints that return organic results, snippets, and “discussions” (e.g., Reddit/Quora text) from external sources at runtime, which the agent then ingests as readable free text (snippets/thread content) into the LLM context—an outsider-authored injection surface.