zergscholar

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

• Third-party content exposure detected (high risk: 0.75). The “pull-paper” workflow reads outsider-authored free text from the zergscholar server (paper fields like abstract and collaborator/AI knowledge entries via /api/knowledge), then renders it into the agent’s LLM-readable context by writing it into the generated Obsidian note body (and thus into subsequent LLM runs that read that note).