NVIDIA-TAOOfficial·60 Agent Skills Included

tao-skill-bank

Train, evaluate, and deploy NVIDIA TAO vision models

Trains, fine-tunes, evaluates, and deploys NVIDIA TAO computer vision models across Docker, Kubernetes, SLURM, and Brev GPU platforms. Removes manual spec writing, container setup, checkpoint tracking, and cluster job management through a shared four-verb execution contract. Covers dozens of models for detection, segmentation, depth, pose, and video, with built-in AutoML, job records, and S3 data staging.

83|20

↗

npx skills add NVIDIA-TAO/tao-skill-bank --all -g -y

Available:

Tells the agent how to route user requests to the right TAO model, data, and platform skills, then launch and monitor GPU jobs through the shared submit/status/logs/cancel contract without any SDK.

All Skills in This Repository (60)

Pure Emerald Level Indicators

📦 In Repo

NVIDIA-TAO

REPLACE-WITH-PLATFORM-NAME

Automate GPU job deployment on REPLACE-PLATFORM via APIs and Docker.

replace-with-skill-name

Convert PDF documents to plain text for editing.

pdf

Extract text and structured data from PDF documents using pypdf and pdfplumber.

model-trainer

Configure environments and execute AI model training workflows.

skill-skeleton

Generate a template for new data transformation skills with metadata and instructions.

tao-launch-workflow

Automate TAO workflow launches with platform selection, credentials, and preflight checks.

tao-list-capabilities

List TAO Skill Bank plugin capabilities for applications, workflows, models, and AutoML.

tao-run-on-kubernetes

Submit TAO container jobs as Kubernetes Jobs with NVIDIA GPU scheduling.

tao-run-on-brev

Automate TAO training, evaluation, and inference jobs on Brev GPU instances via the Brev CLI.

tao-run-platform

Submit and monitor GPU training jobs on NVIDIA TAO platforms.

tao-run-on-local-docker

Execute TAO SDK jobs in Docker containers with NVIDIA GPU acceleration.

tao-run-on-docker

Run Docker containers on GPU hosts with NVIDIA Container Toolkit.

Official

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install tao-skill-bank?▼

Run `npx skills add NVIDIA-TAO/tao-skill-bank --all -g -y` in your terminal to install every model, data, and platform skill globally.

What can I do with the TAO Skill Bank?▼

You can train, fine-tune, evaluate, export, and run inference on NVIDIA TAO vision models such as DINO, RT-DETR, Mask2Former, CLIP, and Cosmos, then deploy them to TensorRT.

Which platforms can run TAO training jobs?▼

Jobs run on local Docker, remote Docker over SSH, Kubernetes, SLURM clusters, Brev GPU instances, or a plain Python virtualenv, all through the same four-verb contract.

Does the TAO Skill Bank require the TAO SDK?▼

No. Job tracking, S3 data staging, and multi-node training are built in and run over each platform's native CLI; only the AutoML workflow uses a small extra wheel.

Do I need coding experience to train a TAO model?▼

No. Describe your dataset and goal in plain language, and the agent resolves the model, writes the spec, checks prerequisites, and launches the GPU job after your confirmation.