perf-parallelism-strategies

Pass

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill serves as a technical reference guide for high-performance computing (HPC) parallelism strategies.
  • [SAFE]: Code examples provided in the markdown are configuration snippets for model parameters (e.g., tensor_model_parallel_size) and do not contain sensitive information or unsafe commands.
  • [SAFE]: The 'Verification' section contains a uv run python command using torch.distributed.run. This is a standard practice for launching distributed training jobs in this domain and targets local scripts (scripts/training/run_recipe.py) with local configuration overrides. It does not download or execute remote code.
  • [SAFE]: External links point to a legitimate academic source (arXiv) for technical documentation.
  • [SAFE]: No obfuscation, prompt injection, or data exfiltration patterns were identified.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 24, 2026, 12:14 AM
Security Audit — agent-trust-hub — perf-parallelism-strategies