# Tao run deft od aoi

> Tao run deft od aoi is an open-source agent skill for design and media. Runs an iterative RT-DETR retraining loop for binary industrial defect detection. It is available on Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose.

- **Kind:** Skill
- **Pricing:** Open source
- **Platforms:** Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
- **Website:** https://github.com/NVIDIA/skills/tree/main/skills/tao-run-deft-od-aoi
- **Repository:** https://github.com/nvidia/skills
- **GitHub stars:** 4,037
- **License:** Apache-2.0
- **Skill name:** tao-run-deft-od-aoi
- **Collection:** https://github.com/NVIDIA/skills
- **Categories:** design and media, computer vision
- **Listing updated:** 2026-11-18

## About

This skill drives a disk-backed training loop for a single foreground class, defect, aimed at automated optical inspection. Each round measures the current model on frozen KPI and test sets, splits errors into false negatives and false positives, and uses a frozen SigLIP encoder to find visually similar real and clean images from pooled sources. Admitted images and boxes accumulate into one binary COCO dataset, and every round retrains from the same base checkpoint, keeping the checkpoint with the best KPI AP50.

The workflow is organized as scripts that prepare source COCO files, initialize a policy, build gap and retrieval specs, admit candidates, and commit stage artifacts to a state file with hashes. It relies on TAO Data Services and RT-DETR containers that the agent submits and polls through the selected platform, so the run needs CUDA GPUs and COCO datasets whose roles are kept disjoint. An optional AnomalyGenNext synthesis route, gated by pixel masks and per-route checkpoints, can add generated defect images.

## When your agent uses it

> Run a binary industrial-inspection DEFT loop with TAO RT-DETR: measure on frozen KPI/test roles, mine gap-similar real and clean images with SigLIP, accumulate admitted COCO data, retrain from one base checkpoint, and select by KPI AP50. Use for iterative AOI defect detection, not generic multiclass OD.

## What it can access

From a static check of the files on GitHub; nothing was run.

- [Runs Python scripts](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-run-deft-od-aoi)
- [Asks for: Read, Bash, Write](https://github.com/NVIDIA/skills/blob/HEAD/skills/tao-run-deft-od-aoi/SKILL.md)
- [No network calls found in its scripts](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-run-deft-od-aoi)
- [License: Apache-2.0](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-run-deft-od-aoi)

## Features

- **Frozen KPI measurement:** Scores each round on held-out KPI and test roles, with test results kept report-only.
- **Gap-based image mining:** Routes false negatives and false positives to real or clean image pools using SigLIP similarity.
- **Cumulative COCO admission:** Adds deduplicated images and boxes each round and caps clean negatives against real defects.
- **Stage commit tracking:** Records each completed stage with file hashes in a durable state file and loop log.
- **Probe-based training selection:** Runs candidate training probes and selects the best KPI checkpoint, with an optional late extension.
- **Optional defect synthesis:** Can generate defect images through AnomalyGenNext for routes that supply masks and checkpoints.

## Good for

- Improve recall of a binary defect detector on a production inspection line
- Mine look-alike real images that a current RT-DETR model misses
- Build a cumulative defect dataset across several retraining rounds
- Check that KPI, test and training image roles stay disjoint before training
- Select the best defect detector checkpoint by KPI AP50

## Install

### Claude Code

Copy the skill folder to `~/.claude/skills/tao-run-deft-od-aoi` to use it in all your projects:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.claude/skills &&
rm -rf ~/.claude/skills/tao-run-deft-od-aoi &&
cp -R "$tmp/skills/tao-run-deft-od-aoi" ~/.claude/skills/tao-run-deft-od-aoi &&
rm -rf "$tmp"
```

For one project only, put the folder in `.claude/skills/tao-run-deft-od-aoi` inside it instead.

### Codex

Copy the skill folder to `~/.agents/skills/tao-run-deft-od-aoi` to use it in every repository:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.agents/skills &&
rm -rf ~/.agents/skills/tao-run-deft-od-aoi &&
cp -R "$tmp/skills/tao-run-deft-od-aoi" ~/.agents/skills/tao-run-deft-od-aoi &&
rm -rf "$tmp"
```

For one repository only, put the folder in `.agents/skills/tao-run-deft-od-aoi` inside it instead.

### Cursor

Copy the skill folder to `~/.cursor/skills/tao-run-deft-od-aoi` to use it in every project:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.cursor/skills &&
rm -rf ~/.cursor/skills/tao-run-deft-od-aoi &&
cp -R "$tmp/skills/tao-run-deft-od-aoi" ~/.cursor/skills/tao-run-deft-od-aoi &&
rm -rf "$tmp"
```

For one project only, put the folder in `.cursor/skills/tao-run-deft-od-aoi` inside it instead.

### Gemini CLI

Run in a terminal (add `--scope workspace` to install it for the current workspace only):

```sh
gemini skills install https://github.com/NVIDIA/skills.git --path skills/tao-run-deft-od-aoi
```

Or copy the folder to `~/.gemini/skills/tao-run-deft-od-aoi` yourself.

### VS Code

For GitHub Copilot in VS Code, copy the skill folder to `~/.copilot/skills/tao-run-deft-od-aoi`:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.copilot/skills &&
rm -rf ~/.copilot/skills/tao-run-deft-od-aoi &&
cp -R "$tmp/skills/tao-run-deft-od-aoi" ~/.copilot/skills/tao-run-deft-od-aoi &&
rm -rf "$tmp"
```

For one project only, put the folder in `.github/skills/tao-run-deft-od-aoi` inside it instead.

### Other agents

Get the folder:

```sh
git clone --depth 1 https://github.com/NVIDIA/skills.git
```

Then copy `skills/skills/tao-run-deft-od-aoi` into your agent's skills folder as `tao-run-deft-od-aoi`. Agents that support SKILL.md read one folder per skill.

## FAQ

### What does this skill need to run?

It requires TAO RT-DETR and Data Services images, CUDA GPUs, and normalized COCO roles for KPI, test, defective-real and verified-clean data.

### Does it work for multiclass object detection?

No. It is built for a single foreground class, defect, and the skill's own description says it is not meant for generic multiclass detection.

### Which checkpoint is kept after training?

The checkpoint with the highest KPI val_mAP50 is carried forward; test results do not influence selection.

### Is synthetic data required?

No. The core loop uses real data only, and AnomalyGenNext synthesis is an optional route with its own gates.

### How do I install Tao run deft od aoi in Claude Code?

Copy the skill folder from GitHub to ~/.claude/skills/tao-run-deft-od-aoi to use it in all your projects, or to .claude/skills/tao-run-deft-od-aoi in one project.

### Does Tao run deft od aoi run scripts?

Yes. It runs Python scripts. No network calls were found in its scripts. This comes from a static check of the files on GitHub; nothing was run. Read the files before you install it.

### Which AI agents can use Tao run deft od aoi?

Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose. Each reads skills in the SKILL.md format from its own folder; the install steps on this page list them.

### What are the best alternatives to Tao run deft od aoi?

The closest alternatives to Tao run deft od aoi are Rea tool design, Prototype, Codebase design, Frontend design and Brand guidelines.

## Where this comes from

- Description: from the skill's own SKILL.md.

## Alternatives

- [Rea tool design](https://ai-powered.com/skills/rea-tool-design): Rea tool design is an open-source agent skill for design and media.
- [Prototype](https://ai-powered.com/skills/prototype): Prototype is an open-source agent skill for developer tools.
- [Codebase design](https://ai-powered.com/skills/codebase-design): Codebase design is an open-source agent skill for design and media.
- [Frontend design](https://ai-powered.com/skills/frontend-design): Frontend design is a free agent skill for design and media.
- [Brand guidelines](https://ai-powered.com/skills/brand-guidelines): Brand guidelines is a free agent skill for design and media.
- [PPTX](https://ai-powered.com/skills/pptx): PPTX is a free agent skill for files and documents.
- [Theme factory](https://ai-powered.com/skills/theme-factory): Theme factory is a free agent skill for design and media.
- [Slack GIF creator](https://ai-powered.com/skills/slack-gif-creator): Slack GIF creator is a free agent skill for communication.

---
Source: https://ai-powered.com/skills/tao-run-deft-od-aoi
