grok-deep-research

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests an outsider-provided research brief via --query or --query-file into the Grok prompts (including the synthesis prompt) without selecting a specific trusted item first, enabling indirect prompt-injection into the web research loop.