tao-train-mask-grounding-dino
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill documentation and action schemas describe standard procedures for using the TAO Toolkit. All external resources, such as container images, are hosted on official NVIDIA registries.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as it processes external dataset files for training and evaluation.
- Ingestion points: Dataset annotation files (COCO/ODVG format) specified in action configurations in SKILL.md and references/skill_info.yaml.
- Boundary markers: Absent in the provided YAML configuration templates.
- Capability inventory: Shell command execution for TAO training, evaluation, and TensorRT deployment within controlled Docker environments.
- Sanitization: Not explicitly defined; the skill assumes the datasets follow the expected structured JSON/JSONL schemas.
Audit Metadata