Skills
skills/ascend-ai-coding/awesome-ascend-skills/mindspeed-mm-train-profiler/Gen Agent Trust Hub

mindspeed-mm-train-profiler

Pass

Audited by Gen Agent Trust Hub on Jun 15, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several local commands to manage the profiling environment and execution process:
  • Environment Setup: Runs source commands for the Ascend toolkit/CANN environment variables (/usr/local/Ascend/ascend-toolkit/set_env.sh or /usr/local/Ascend/cann/set_env.sh).
  • Hardware/Software Checks: Executes npu-smi info to check NPU status and a Python snippet to verify torch and torch_npu availability.
  • Training Execution: Executes a shell script provided by the user (TRAIN_SCRIPT) after creating a modified copy to include profiling parameters.
  • Profiling Tools: Invokes local Python scripts (mindspeed_mm/tools/profiler.py or mindspeed_mm/fsdp/tools/profiler.py) for data parsing and analysis.
  • File System Operations: Uses mkdir -p to create output directories for profiling data.
  • [SAFE]: The skill implements best practices to protect user data by explicitly prohibiting the modification of original user scripts or configuration files, instead requiring the creation of timestamped copies for profiling tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 15, 2026, 06:59 AM
Security Audit — agent-trust-hub — mindspeed-mm-train-profiler