torch-npu-comm-test

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for benchmarking Ascend NPU communication performance using the standard torch.distributed interface. The implementation uses legitimate libraries (torch, torch_npu) for its stated purpose.
  • [SAFE]: Command execution is restricted to the use of torchrun to orchestrate distributed processes, which is the standard and expected method for launching PyTorch distributed workloads.
  • [SAFE]: No evidence of data exfiltration, hardcoded credentials, or unauthorized network activity was found. Network operations are inherent to the skill's purpose (distributed communication testing) and use established protocols (HCCL).
  • [SAFE]: The code does not employ obfuscation, persistence mechanisms, or dynamic execution of untrusted content. Environment setup instructions involve sourcing standard local configuration files for the Ascend toolkit.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 09:32 AM
Security Audit — agent-trust-hub — torch-npu-comm-test