ml-matgl-finetune
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill does not contain any detected malicious instructions, obfuscation, or persistence mechanisms. All scripts perform legitimate data processing and machine learning tasks.\n- [COMMAND_EXECUTION]: The documentation and skill instructions direct the agent to execute Python scripts (prepare_matgl_data.py, train_matgl.py) using conda run. These executions are limited to the skill's own internal logic for processing datasets and managing the training loop.\n- [EXTERNAL_DOWNLOADS]: The training script utilizes matgl.load_model to fetch pre-trained foundation models (e.g., CHGNet). These downloads are performed from the official MatGL model registry, which is a trusted and standard source for materials science research.\n- [DATA_INGESTION]: The prepare_matgl_data.py script ingests user-supplied JSON files containing atomic structures and labels. The ingestion is restricted to numeric data extraction and standard unit conversions, presenting no significant surface for indirect prompt injection.
Audit Metadata