ultralytics

Plugin ultralytics from marketplace ultralytics-plugins in Bovey0809/claude-code-ultralytics. Commands read from the repo's manifest — 1 month ago.

Skills (4)

• analyze

  Use when the user wants to analyze, debug, or explain a YOLO model's training, validation, dataset, or predictions — questions like "why is mAP low", "is my dataset balanced", "did this overfit", "what classes does it confuse", "compare these two runs", "analyze loss", "find failure cases". Dispatch

• modelbox-create-flowunit

  Use when authoring a NEW custom ModelBox flowunit — Python, C++, or an inference (model-wrapper) flowunit — and wiring it into a graph. Generates the flowunit directory (`<name>.toml` plus code/CMake), wires it to a graph .toml, and validates the layout. Use when the user asks to "create a flowunit"

• modelbox

  Use when authoring or modifying a ModelBox graph (.toml), configuring a ModelBox flowunit, choosing the right flowunit for a step (video decode, inference, YOLO postprocess, drawing boxes, encoding output), or resolving "which flowunit handles X". Provides a curated index of flowunits used in YOLO +

• yolo

  Use when the user mentions Ultralytics or YOLO — training a detector/segmenter/classifier/pose/OBB model, dataset YAML files, .pt weights, predicting on images/video, or exporting to ONNX/CoreML/TensorRT/TFLite/OpenVINO. Provides the conceptual model and dispatches to the matching slash command (/tr

Slash commands (4)

• /ultralytics:dataset-check

  Validate an Ultralytics dataset YAML — checks paths resolve, splits have images, labels exist, class ids match `names`. Pure validation, does not invoke yolo.

• /ultralytics:export

  Export a YOLO .pt model to ONNX, CoreML, TensorRT, TFLite, OpenVINO, etc.

• /ultralytics:predict

  Run YOLO inference on an image, directory, video, URL, or webcam. Accepts key=value args or natural language.

• /ultralytics:train

  Train an Ultralytics YOLO model. Accepts key=value args (model=, data=, epochs=, …) or natural language.

Similar plugins