getitune-training-a-model

Pass

Audited by Gen Agent Trust Hub on Jul 23, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the library's official CLI tool getitune for training, testing, and prediction tasks. It also references the just command runner for standard development workflows like linting and unit testing. These operations are restricted to the local development environment.
  • [EXTERNAL_DOWNLOADS]: The documentation includes instructions to use uv sync for environment setup and dependency management. This is a well-known, standard practice for Python project synchronization.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface by instructing the agent to process external datasets and checkpoint files.
  • Ingestion points: Dataset root directories (COCO, YOLO, VOC formats) and model weight files (.pt checkpoints).
  • Boundary markers: None identified in the prompt templates.
  • Capability inventory: Subprocess execution via getitune CLI and file system writes to the specified work_dir.
  • Sanitization: No specific sanitization or validation of the dataset content or checkpoint integrity is described in the instructions.
  • Assessment: While this is a standard requirement for ML training skills, processing untrusted data remains an inherent risk factor.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 23, 2026, 12:21 AM
Security Audit — agent-trust-hub — getitune-training-a-model