academic-pipeline

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text enters at the skill’s runtime via the user’s chat “first message” / stage inputs, and the orchestrator then passes that dialogue into downstream agents (e.g., collaboration_depth_agent reads raw dialogue turns; integrity/audit gates also consume the user-provided manuscript text and claims at checkpoints), so arbitrary user text can be ingested without selecting a specific item first.