research

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

• Third-party content exposure detected (high risk: 0.90). The workflow (Step 4: "Dispatch Agents") explicitly dispatches a discourse-scanner that reads subreddits (public, user-generated content) and ingests those findings into the synthesis/ranking/report process, so untrusted third‑party forum content can influence decisions and actions.