physical-ai-physics-selection

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides detailed decision-making logic for choosing simulation stacks based on task requirements (locomotion, grasping, etc.) without any malicious instructions.\n- [COMMAND_EXECUTION]: The script scripts/probe_environment.py is a benign utility that uses importlib.metadata to inventory installed packages related to the physics simulation stack. It does not initialize CUDA or execute untrusted logic.\n- [EXTERNAL_DOWNLOADS]: The skill references and provides installation commands for NVIDIA-owned resources including the ovnewton and newton packages. These downloads are from trusted vendor sources and are essential for the skill's stated purpose.\n- [INDIRECT_PROMPT_INJECTION]: While the skill processes user project requirements, it includes instructions to separate vendor claims from measured evidence and verify task behavior, which serves as a defensive boundary.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 03:52 PM