tao-finetune-kumo-anomaly

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches source code from the official NVIDIA repository on GitHub (github.com/NVIDIA/Kumo-TS).
  • [EXTERNAL_DOWNLOADS]: Downloads pretrained model weights and configuration files from the official NVIDIA organization on Hugging Face (huggingface.co/nvidia/Kumo-Anomaly).
  • [COMMAND_EXECUTION]: Provides instructions to use the uv package manager and git for environment setup and repository cloning.
  • [COMMAND_EXECUTION]: Includes examples for running Python scripts (quick_example.py, finetune_example.py) to perform inference and model training using local data.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data via CSV files which acts as an ingestion surface.
  • Ingestion points: CSV files passed to --dataset-path or --csv arguments in SKILL.md.
  • Boundary markers: Not explicitly defined for data delimiters.
  • Capability inventory: The skill performs file system reads/writes and subprocess execution via Python scripts.
  • Sanitization: Input validation is mentioned where the API raises a ValueError if non-numeric columns are detected.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 04:07 PM