ml-fairchem-finetune
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell commands within
run.shandtest_fairchem_finetuning.shto automate the fine-tuning pipeline. This includes executingpythonscripts and thefairchemCLI to perform data preparation and model training. - [SAFE]: The skill dynamically generates Python helper scripts (
_freeze_backbone_helper.py) and Hydra configuration YAML files. These are used to customize the training runner and are generated from hardcoded templates within the skill, which is a standard approach for experiment management in the Fairchem ecosystem. - [PROMPT_INJECTION]: An indirect prompt injection surface exists as the skill processes user-supplied JSON data files through
scripts/prepare_fairchem_data.py. However, the ingested data is strictly used for numerical model training and data conversion, and is not used to influence or construct LLM prompts, effectively mitigating the risk.
Audit Metadata