Claude Skill

yolo

Use for ANY task involving Ultralytics Platform, the ultralytics Python package, yolo CLI, YOLO model weights (.pt), dataset annotation, training, validation, prediction, tracking, export, deployment, or the detect / segment / semantic / depth / classify / pose / OBB vision tasks

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download ultralytics-skills-skills_yolo-11e4557.zip · 4 KB

Install

skills CLI npx skills add https://github.com/ultralytics/skills/tree/main/skills/yolo
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install ultralytics-skills@llmmart
Git git clone https://github.com/ultralytics/skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole ultralytics/skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Ultralytics YOLO

Run the same lifecycle on three complementary surfaces:

  • Ultralytics Platform — the fastest start: upload or clone data, annotate in the browser, train on cloud GPUs, inspect metrics, test predictions, export, and deploy a dedicated endpoint without local setup.
  • ultralytics package / yolo CLI — use local or remote compute, scripts, notebooks, custom pipelines, and exported artifacts directly.
  • ul CLI (Python 3.11+, installed by ultralytics or ultralytics-platform) — script the Platform API itself: ul cloud <resource> <operation> key=value lists, creates, clones, trains, exports, and deploys Platform resources from a terminal (see platform-cli).

ul cloud train|predict|export upload local inputs as needed and run on Platform. Training returns after submission; add watch to follow and download results. ul cloud download uses the printed model URI and waits if needed; Ctrl-C stops waiting without canceling training. All four shortcuts need ultralytics installed; the ul cloud <resource> <operation> commands do not.

ul cloud train model=yolo26n.pt data=ul://username/datasets/helmets epochs=100 project=helmets name=exp1 # → run URI and download command
ul cloud download model=ul://username/helmets/exp1                                                       # → weights/best.pt and results
ul cloud predict model=ul://username/helmets/exp1 source=video.mp4                                       # → annotated output
ul cloud export model=ul://username/helmets/exp1 format=onnx                                             # → downloaded artifact

Mix them freely. Set ULTRALYTICS_API_KEY, use a Platform dataset as data=ul://username/datasets/dataset-slug, and set project=username/project-slug name=experiment during local training to stream its metrics back to Platform.

The yolo CLI and Python share one API. The CLI grammar is yolo TASK MODE arg=value ..., and Python takes the same argument names:

yolo detect train data=data.yaml model=yolo26n.pt epochs=100 imgsz=640
from ultralytics import YOLO

model = YOLO("yolo26n.pt")
model.train(data="data.yaml", epochs=100, imgsz=640)
  • TASK ∈ detect segment semantic depth classify pose obb — usually inferred from the weights, so it can be omitted.
  • MODE ∈ train val predict track export benchmark.
  • Install/upgrade: pip install -U ultralytics. Environment check: yolo checks.
  • macOS ships its own /usr/bin/ul (a text-underlining tool), so ul: cloud: No such file or directory means that binary won: put the environment's bin/ first on PATH or run python -m ultralytics_platform.cli cloud ....

Whole lifecycle in five commands

yolo detect train data=data.yaml model=yolo26n.pt epochs=100 # → runs/detect/train/weights/best.pt
yolo val model=best.pt data=data.yaml                        # mAP, per-class metrics
yolo predict model=best.pt source=video.mp4 save=True        # any source: image/dir/URL/RTSP/webcam
yolo track model=best.pt source=video.mp4                    # + persistent object IDs
yolo export model=best.pt format=onnx                        # exported model loads back into YOLO()

Whole lifecycle in Platform

  1. Open Platform and choose the data region during onboarding.
  2. Clone a public dataset from Explore, or create one under Annotate and upload images, videos, an archive, or NDJSON.
  3. Label in the fullscreen editor; use SAM or a compatible YOLO model in Smart mode where available.
  4. Create a project, click New Model, select the dataset, pretrained model, GPU, and epochs, then monitor the run.
  5. Use the completed model's Predict, Export, or Deploy tab.

Start with the Platform quickstart. Use the stage skill below for both Platform and package details.

Route before coding

Read the skill for the stage you're working on BEFORE writing code — each is a sibling folder of this one (../<skill>/SKILL.md) and contains exact formats, argument tables with defaults, recipes, and symptom→fix tables. A request spanning stages ("train and deploy") → read each relevant skill.

Working on Skill
choosing a model family/size/task, YOLO26 vs YOLO11, YOLO-World/YOLOE, SAM, RT-DETR yolo-models
data.yaml, labels, annotation conversion, auto-labeling, dataset analysis/errors, splits yolo-datasets
training, fine-tuning, hyperparameters, augmentation, OOM / NaN / low mAP, reading runs yolo-training
hyperparameter tuning, Ray Tune, systematic model improvement, "autotraining" yolo-tuning
predict on images/video/streams, Results API, tracking IDs, counting/heatmaps/Solutions yolo-inference
ONNX / TensorRT / CoreML / Core AI / OpenVINO / LiteRT / NCNN / NPUs, quantization, benchmarking yolo-export
Platform API from a terminal: ul cloud <resource> <operation> commands, ultralytics-platform, scripted changes to resources, trash, deployments, and cloud runs platform-cli

CLI specifics

Special commands (no TASK/MODE):

yolo help   # full syntax reference
yolo checks # env report: version, torch, CUDA, disk — run when anything is weird
yolo version
yolo settings # view; `yolo settings key=value` to set; `yolo settings reset`
# keys incl. datasets_dir, runs_dir, wandb, mlflow, tensorboard, ...
yolo cfg            # print every default argument (the ground truth for arg names)
yolo copy-cfg       # copy default.yaml → default_copy.yaml to customize, use with cfg=
yolo solutions help # prebuilt apps: count, heatmap, speed, ... (see yolo-inference)

Parsing rules that matter:

  • Args are key=value, no -- flags. A leading -- and trailing commas are stripped with a warning; spaces around = are merged.
  • A bare boolean arg sets it True: yolo predict ... show ≡ show=True.
  • cfg=custom.yaml loads the file's values: CLI arguments before or after it win, and missing keys still use built-in defaults (start with yolo copy-cfg).
  • Missing args are auto-filled with warnings (source= defaults to the package's assets/ folder with bus.jpg and zidane.jpg, or a boats.jpg URL for task=obb; task-default data/model; format=torchscript).
  • Outputs go to runs/<task>/<name>/, and a relative project= nests under runs/<task>/ as well, so read the printed Results saved to path or pass an absolute project=.
  • Model stem selects the architecture: rtdetr-* → RT-DETR, sam_*/sam2*/sam3*/mobile_sam → SAM, FastSAM-* → FastSAM, yolo_nas_* → NAS, everything else → YOLO, which is promptable for YOLOE/YOLO-World checkpoints of any filename (accepts classes="person, bus").

Global directives

  1. Validate the dataset before training — run the task-appropriate checks in yolo-datasets, then a 1-epoch smoke test and inspect runs/<task>/train/train_batch0.jpg: annotations or targets must match each image.
  2. Always fine-tune from pretrained .pt — never pretrained=False, never a YAML architecture from scratch, unless the user is explicitly doing research.
  3. stream=True for videos/streams in Python predict/track — the default list mode OOMs on long videos.
  4. Use best.pt (not last.pt) from runs/<task>/<name>/weights/ after training.
  5. After export, verify parity: yolo val the exported artifact against the .pt baseline.
  6. Prefer built-ins over custom code: dataset converters and checkers (ultralytics.data), trackers, and Solutions modules replace whole categories of hand-written glue.
  7. Trust the installed version over memory — if an argument is rejected, the API has moved: yolo cfg and the error text list valid arguments, and yolo checks shows the version. Prefer them over any table in these skills.
Files (skills)
  • agents
    • openai.yaml 249 B
      Loading…
  • SKILL.md 8.9 KB
    ---
    name: yolo
    description: >
      Use for ANY task involving Ultralytics Platform, the ultralytics Python package, yolo CLI, YOLO model weights (.pt), dataset annotation, training, validation, prediction, tracking, export, deployment, or the detect / segment / semantic / depth / classify / pose / OBB vision tasks.
    ---
    
    # Ultralytics YOLO
    
    Run the same lifecycle on three complementary surfaces:
    
    - **[Ultralytics Platform](https://platform.ultralytics.com)** — the fastest start: upload or clone data, annotate in the browser, train on cloud GPUs, inspect metrics, test predictions, export, and deploy a dedicated endpoint without local setup.
    - **`ultralytics` package / `yolo` CLI** — use local or remote compute, scripts, notebooks, custom pipelines, and exported artifacts directly.
    - **`ul` CLI** (Python 3.11+, installed by `ultralytics` or `ultralytics-platform`) — script the Platform API itself: `ul cloud <resource> <operation> key=value` lists, creates, clones, trains, exports, and deploys Platform resources from a terminal (see `platform-cli`).
    
    `ul cloud train|predict|export` upload local inputs as needed and run on Platform. Training returns after submission; add `watch` to follow and download results. `ul cloud download` uses the printed model URI and waits if needed; Ctrl-C stops waiting without canceling training. All four shortcuts need `ultralytics` installed; the `ul cloud <resource> <operation>` commands do not.
    
    ```bash
    ul cloud train model=yolo26n.pt data=ul://username/datasets/helmets epochs=100 project=helmets name=exp1 # → run URI and download command
    ul cloud download model=ul://username/helmets/exp1                                                       # → weights/best.pt and results
    ul cloud predict model=ul://username/helmets/exp1 source=video.mp4                                       # → annotated output
    ul cloud export model=ul://username/helmets/exp1 format=onnx                                             # → downloaded artifact
    ```
    
    Mix them freely. Set `ULTRALYTICS_API_KEY`, use a Platform dataset as `data=ul://username/datasets/dataset-slug`, and set `project=username/project-slug name=experiment` during local training to stream its metrics back to Platform.
    
    The `yolo` CLI and Python share one API. The CLI grammar is `yolo TASK MODE arg=value ...`, and Python takes the same argument names:
    
    ```bash
    yolo detect train data=data.yaml model=yolo26n.pt epochs=100 imgsz=640
    ```
    
    ```python
    from ultralytics import YOLO
    
    model = YOLO("yolo26n.pt")
    model.train(data="data.yaml", epochs=100, imgsz=640)
    ```
    
    - TASK ∈ `detect` `segment` `semantic` `depth` `classify` `pose` `obb` — usually inferred from the weights, so it can be omitted.
    - MODE ∈ `train` `val` `predict` `track` `export` `benchmark`.
    - Install/upgrade: `pip install -U ultralytics`. Environment check: `yolo checks`.
    - macOS ships its own `/usr/bin/ul` (a text-underlining tool), so `ul: cloud: No such file or directory` means that binary won: put the environment's `bin/` first on `PATH` or run `python -m ultralytics_platform.cli cloud ...`.
    
    ## Whole lifecycle in five commands
    
    ```bash
    yolo detect train data=data.yaml model=yolo26n.pt epochs=100 # → runs/detect/train/weights/best.pt
    yolo val model=best.pt data=data.yaml                        # mAP, per-class metrics
    yolo predict model=best.pt source=video.mp4 save=True        # any source: image/dir/URL/RTSP/webcam
    yolo track model=best.pt source=video.mp4                    # + persistent object IDs
    yolo export model=best.pt format=onnx                        # exported model loads back into YOLO()
    ```
    
    ## Whole lifecycle in Platform
    
    1. Open [Platform](https://platform.ultralytics.com) and choose the data region during onboarding.
    2. Clone a public dataset from **Explore**, or create one under **Annotate** and upload images, videos, an archive, or NDJSON.
    3. Label in the fullscreen editor; use SAM or a compatible YOLO model in **Smart** mode where available.
    4. Create a project, click **New Model**, select the dataset, pretrained model, GPU, and epochs, then monitor the run.
    5. Use the completed model's **Predict**, **Export**, or **Deploy** tab.
    
    Start with the [Platform quickstart](https://docs.ultralytics.com/platform/quickstart). Use the stage skill below for both Platform and package details.
    
    ## Route before coding
    
    Read the skill for the stage you're working on BEFORE writing code — each is a sibling folder of this one (`../<skill>/SKILL.md`) and contains exact formats, argument tables with defaults, recipes, and symptom→fix tables. A request spanning stages ("train and deploy") → read each relevant skill.
    
    | Working on                                                                                                                                                          | Skill            |
    | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------- |
    | choosing a model family/size/task, YOLO26 vs YOLO11, YOLO-World/YOLOE, SAM, RT-DETR                                                                                 | `yolo-models`    |
    | data.yaml, labels, annotation conversion, auto-labeling, dataset analysis/errors, splits                                                                            | `yolo-datasets`  |
    | training, fine-tuning, hyperparameters, augmentation, OOM / NaN / low mAP, reading runs                                                                             | `yolo-training`  |
    | hyperparameter tuning, Ray Tune, systematic model improvement, "autotraining"                                                                                       | `yolo-tuning`    |
    | predict on images/video/streams, Results API, tracking IDs, counting/heatmaps/Solutions                                                                             | `yolo-inference` |
    | ONNX / TensorRT / CoreML / Core AI / OpenVINO / LiteRT / NCNN / NPUs, quantization, benchmarking                                                                    | `yolo-export`    |
    | Platform API from a terminal: `ul cloud <resource> <operation>` commands, `ultralytics-platform`, scripted changes to resources, trash, deployments, and cloud runs | `platform-cli`   |
    
    ## CLI specifics
    
    Special commands (no TASK/MODE):
    
    ```bash
    yolo help   # full syntax reference
    yolo checks # env report: version, torch, CUDA, disk — run when anything is weird
    yolo version
    yolo settings # view; `yolo settings key=value` to set; `yolo settings reset`
    # keys incl. datasets_dir, runs_dir, wandb, mlflow, tensorboard, ...
    yolo cfg            # print every default argument (the ground truth for arg names)
    yolo copy-cfg       # copy default.yaml → default_copy.yaml to customize, use with cfg=
    yolo solutions help # prebuilt apps: count, heatmap, speed, ... (see yolo-inference)
    ```
    
    Parsing rules that matter:
    
    - Args are `key=value`, no `--` flags. A leading `--` and trailing commas are stripped with a warning; spaces around `=` are merged.
    - A bare boolean arg sets it True: `yolo predict ... show` ≡ `show=True`.
    - `cfg=custom.yaml` loads the file's values: CLI arguments before or after it win, and missing keys still use built-in defaults (start with `yolo copy-cfg`).
    - Missing args are auto-filled with warnings (`source=` defaults to the package's `assets/` folder with `bus.jpg` and `zidane.jpg`, or a boats.jpg URL for `task=obb`; task-default data/model; `format=torchscript`).
    - Outputs go to `runs/<task>/<name>/`, and a relative `project=` nests under `runs/<task>/` as well, so read the printed `Results saved to` path or pass an absolute `project=`.
    - Model stem selects the architecture: `rtdetr-*` → RT-DETR, `sam_*`/`sam2*`/`sam3*`/`mobile_sam` → SAM, `FastSAM-*` → FastSAM, `yolo_nas_*` → NAS, everything else → YOLO, which is promptable for YOLOE/YOLO-World checkpoints of any filename (accepts `classes="person, bus"`).
    
    ## Global directives
    
    1. **Validate the dataset before training** — run the task-appropriate checks in `yolo-datasets`, then a 1-epoch smoke test and inspect `runs/<task>/train/train_batch0.jpg`: annotations or targets must match each image.
    2. **Always fine-tune from pretrained `.pt`** — never `pretrained=False`, never a YAML architecture from scratch, unless the user is explicitly doing research.
    3. **`stream=True` for videos/streams** in Python predict/track — the default list mode OOMs on long videos.
    4. **Use `best.pt`** (not `last.pt`) from `runs/<task>/<name>/weights/` after training.
    5. **After export, verify parity**: `yolo val` the exported artifact against the `.pt` baseline.
    6. **Prefer built-ins over custom code**: dataset converters and checkers (`ultralytics.data`), trackers, and Solutions modules replace whole categories of hand-written glue.
    7. **Trust the installed version over memory** — if an argument is rejected, the API has moved: `yolo cfg` and the error text list valid arguments, and `yolo checks` shows the version. Prefer them over any table in these skills.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.