deep-research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Research step explicitly instructs the agent to "Run multiple web searches" and ingest content from open/public sources including "Reddit/HN/Twitter discussions" and company blogs, meaning the agent will fetch and interpret untrusted third-party user-generated content that can influence its analysis and actions.