iblai-vibe-agent-evals

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

• Third-party content exposure detected (medium risk: 0.30). In the AgentEvaluationTab runtime workflow, tenant users can upload/manage benchmark Q&A items (including via CSV) and then trigger background evaluation runs and LLM-as-judge reviews, so outsider-authored free text from those benchmark items (and trace inputs/outputs) is ingested into the evaluation/judge logic without any intermediate trusted selection step.