cli-anything-unimol-tools

Pass

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references external GitHub repositories (such as the dptech-corp Uni-Mol repository and a user-hosted example repository) for datasets, documentation, and scientific references.\n- [COMMAND_EXECUTION]: Defines a CLI interface via the cli_anything.unimol_tools Python module for running molecular property prediction workflows.\n- [DATA_EXFILTRATION]: The skill's primary functionality involves reading local CSV files containing SMILES strings and target values for training and prediction.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the ingestion of external molecular data files.\n
  • Ingestion points: CSV files loaded through the --data-path argument in the train and predict commands.\n
  • Boundary markers: None explicitly defined in the skill instructions for data processing.\n
  • Capability inventory: Executes Python-based training, evaluation, and prediction tasks.\n
  • Sanitization: Standard data validation is expected within the Uni-Mol Tools backend, but sanitization of potentially malicious instructions in CSV metadata or fields is not mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 18, 2026, 12:02 PM