pai-benchmark-policy
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill documentation provides legitimate usage patterns for the Physical AI Studio benchmarking tools, including the
physicalaiCLI and associated Python library. - [EXTERNAL_DOWNLOADS]: The instructions reference configuration examples hosted in the official GitHub repository for the Physical AI Studio. This is a recognized resource belonging to the skill's authoring organization.
- [COMMAND_EXECUTION]: The skill includes instructions for running the
physicalai benchmarkcommand. These are standard entry points for the vendor's platform and do not involve shell injection or privilege escalation attempts. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing benchmark configurations (
.yaml) and model checkpoints (.ckpt). While this represents a data ingestion surface, it is a standard part of the benchmarking process for local artifacts, and no unsafe interpolation or sanitization bypasses are suggested.
Audit Metadata