getitune-preparing-datasets
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for importing and processing external dataset formats (COCO, YOLO, VOC, Datumaro). This creates a potential surface where malicious instructions could be embedded within the dataset's metadata or annotations. \n
- Ingestion points: The
data=parameter in the Python API and--data_rootin the CLI as described inSKILL.md. \n - Boundary markers: None are defined to separate dataset content from agent instructions. \n
- Capability inventory: The skill utilizes file system access and initiates model training via
engine.train(). \n - Sanitization: No explicit sanitization or validation steps for dataset content are mentioned in the instructions.\n- [EXTERNAL_DOWNLOADS]: The skill references the Datumaro library hosted on GitHub. This repository is maintained by the skill author's organization (open-edge-platform) and is a standard dependency for the described functionality.
Audit Metadata