tao-run-deft-od-aoi
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs legitimate industrial inspection workflows using established NVIDIA TAO tools.
- [COMMAND_EXECUTION]: Subprocess calls are used safely to invoke the
annotationsCLI tool for data merging. These calls use argument lists rather than shell strings, mitigating command injection risks. - [INDIRECT_PROMPT_INJECTION]: The skill processes structured data (COCO JSON). It includes robust validation logic to ensure input files match the expected schema and constraints, reducing the surface for indirect injection attacks.
- [DATA_EXPOSURE]: No hardcoded credentials or sensitive file access patterns were detected. The skill manages project-specific data within user-defined directories.
- [OBFUSCATION]: No obfuscated code or hidden instructions were found in the skill's markdown or script files.
Audit Metadata