tao-train-pointpillars
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues detected. The skill is authored by a known vendor and utilizes official resources.
- [EXTERNAL_DOWNLOADS]: The skill documentation references official NVIDIA Docker images from the NVIDIA Container Registry (nvcr.io). These are well-known, vendor-owned services used for distributing the TAO toolkit software.
- [COMMAND_EXECUTION]: The skill provides templates for running model training and inference commands. These commands are restricted to the PointPillars CLI within a Docker environment, which is the intended functional purpose of the skill.
- [PROMPT_INJECTION]: The skill contains logic to guide the agent in determining when to enable AutoML based on user input (e.g., 'turn off AutoML'), which is functional instructional guidance rather than an attempt to bypass security filters.
Audit Metadata