tao-train-grounding-dino

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

• Third-party content exposure detected (medium risk: 0.30). In the required inference runtime, the agent-consumable free text comes from the outsider-provided caption list wired to dataset.infer_data_sources.captions (used as text prompts) and routed into grounding_dino inference -e {config_path} without any requirement to preselect trusted items first.