mindspeed-mm-generative

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download of source code, model weights, and pre-built binaries from official and trusted repositories, including Huawei's GitCode, NVIDIA's GitHub, and the HuggingFace platform.
  • [COMMAND_EXECUTION]: Includes instructions for standard environment setup and model training operations such as repository cloning, system dependency installation, and the execution of training scripts via torchrun.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined configuration files (data.json, data.txt) to manage dataset paths and model parameters. This is a common surface for data ingestion in machine learning workflows, and the skill implements standard usage patterns for local training pipelines.
  • [REMOTE_CODE_EXECUTION]: Software is installed from verified official sources and public package registries (PyPI), following standard development practices for the targeted hardware platform.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 06:27 AM
Security Audit — agent-trust-hub — mindspeed-mm-generative