isaac-cable-insertion-policy-deployment
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Python's
subprocessmodule to orchestrate the deployment workflow. This includes managing Docker containers (create, start, stop, exec), verifying hardware access vianvidia-smi, and performing package integrity checks within the simulation environment. - [EXTERNAL_DOWNLOADS]: It fetches simulation assets (USD files) and container images from official NVIDIA registries (
nvcr.io) and production S3 buckets. All remote assets are validated against hardcoded SHA256 hashes defined in the skill's security boundary logic to prevent tampering. - [DYNAMIC_EXECUTION]: The skill modifies the simulation scene at runtime by injecting a controller script into an in-memory ScriptNode. This mechanism is protected by integrity checks that verify the hashes of both the original scene asset and the controller script before execution is enabled in the simulator.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided machine learning checkpoints (
.pt) and configuration files (.yaml). It includes explicit warnings that checkpoints should only be loaded from trusted sources, as the PyTorch loading process can lead to arbitrary code execution if the input is malicious.
Audit Metadata