The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided datasets (images, labels, and YAML configuration files) across several tasks such as detection, segmentation, and pose estimation.
Ingestion points: Files located at paths specified in data.yaml (e.g., train: images/train), remote dataset URIs (ul://), and local conversion scripts.
Boundary markers: None explicitly defined for the dataset content itself, as it is standard for machine learning tools to process raw data.
Capability inventory: The skill uses the ultralytics Python package and the yolo CLI for training, conversion (convert_coco), and validation (check_det_dataset).
Sanitization: Validation steps are explicitly recommended (check_det_dataset, visualize_image_annotations) to ensure data integrity and format compliance before training begins.
[COMMAND_EXECUTION]: The documentation provides examples of the yolo CLI tool for training models (e.g., yolo train model=yolo26n.pt data=ul://...). These are standard operational commands for the Ultralytics environment.
[EXTERNAL_DOWNLOADS]: The skill mentions the automatic download of standard reference datasets (e.g., coco8.yaml, cityscapes8.yaml) when referenced in the YOLO configuration. These downloads are performed by the official ultralytics library from their established infrastructure.

yolo-datasets