autoresearch

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md Research Loop explicitly runs web searches and WebFetches pages ("For each angle: run 2-3 WebSearch queries" and "For top 2-3 results per angle: WebFetch the page"), which means the agent ingests open/public third‑party content (webpages, social/forum results) and uses that content to drive subsequent research, synthesis, and wiki-filing decisions—exposing it to potential indirect prompt injection.