problem-analysis

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

• Third-party content exposure detected (high risk: 0.85). This skill’s required workflow explicitly instructs runtime research by scraping/reading outsider-authored public sources (e.g., “Reddit, forums, Quora, review sites” and “Capture exact language”), which would be ingested as free-form text into the agent’s LLM context.