tao-run-deft-object-detection agent skill by NVIDIA
Runs the DEFT data augmentation loop for Grounding DINO object detection, with baseline inference, gap analysis, unique-neighbor mining, retraining, and source pool preparation.
At a glance
| Publisher | NVIDIA Official |
|---|---|
| Repository | NVIDIA/skills (396 listed skills) |
| Category | AI Agents and LLM Engineering |
| License | Apache-2.0 |
| Last commit | October 17, 2026 (repository) |
| SKILL.md | 155 lines, 1,786 words |
| Bundled files | 55 files, including references, scripts, assets or templates |
| Quality score | 92/100 · A Passes all seven graded checks |
| Security scan | Not scanned |
| README badge | Get the badges for this listing |
Before you install
Read from the skill's files at the directory's October 19, 2026 snapshot: what ships with it, what its code reads, and the terms it is published under.
- Bundled scripts: 27 script files (25 Python, 2 Shell). Your agent may run them, so read them first.
- Network: its scripts use an HTTP client or a cloud SDK, and name
huggingface.co. - Tool permissions requested:
Read Skill Bash Write(theallowed-toolsfield of SKILL.md). - License: Apache-2.0, an open-source or open-content license.
Install tao-run-deft-object-detection
The skills CLI works with Claude Code, Codex, Cursor and other agents; the other routes copy the folder by hand.
npx skills add https://github.com/NVIDIA/skills/tree/main/skills/tao-run-deft-object-detectiontmp="$(mktemp -d)"
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp/repo"
mkdir -p .claude/skills
cp -R "$tmp/repo/"skills/tao-run-deft-object-detection .claude/skills/
rm -rf "$tmp"python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \
--repo NVIDIA/skills \
--path skills/tao-run-deft-object-detectionNVIDIA/skills is third-party code. Read SKILL.md and any bundled scripts before you install it.
How the author describes it
Run the full DEFT smart-data-augmentation loop for NVIDIA TAO Grounding DINO object detection: zero-shot baseline inference, KPI analysis, per-class gap analysis, SigLIP embedding of weak images, unique-neighbor mining against a source pool, ODVG dataset staging, and retraining — repeated for a fixed number of iterations. Also prepares the source pool the loop mines from, as a separate run: Co-DETR pseudo-labeling, folding to the target classes, KITTI→COCO→ODVG conversion, and embedding. Use for prompts like "run the DEFT OD loop", "run smart data augmentation for grounding dino", "mine and retrain my detection model", "improve OD mAP with gap analysis and mining", "prep the source pool", or "pseudo-label my unlabeled images for mining"; do not use for standalone TAO training, one-off inference, or gap analysis alone.
From the description field of SKILL.md. Agents read this line to decide when to load the skill.
Compatibility, from SKILL.md
| Compatibility | Requires docker + nvidia-container-toolkit and one or more CUDA GPUs. Workflows declare additional requirements. |
|---|
Sections in SKILL.md
- Execution Contract
- Context Discipline
- When to Use This Skill
- Scope: Grounding DINO + ODVG
- Two Invocations: Prep, Then Loop
- Launch Intake
- Agent Behavior
- Workflow
- Using Bundled Scripts
- Stage Reference Modules
- Data, Pre-Flight, Pipeline, and State References
- Gating
Files in the skill folder
BENCHMARK.mdSKILL.mdagents/reporter.mdassets/gap_analysis_object_detection.yamlassets/image_embeddings.yamlassets/overlays/coco_to_odvg.yamlassets/overlays/codetr_inference.yamlassets/overlays/grounding_dino_inference.yamlassets/overlays/kitti_to_coco.yamlassets/overlays/kpi_analyze.yamlassets/tmm_unique_neighbor_matching.yamlassets/train_grounding_dino.yamlconfig/skillspector-baseline.yamlevals/evals.jsonreferences/data-layout.mdreferences/example_classes_its.yamlreferences/grounding-dino.mdreferences/pipeline-and-state.mdreferences/preflight.mdreferences/prep-source-pool.mdreferences/scripts-and-agents.mdreferences/stage-mined-data.mdreferences/tao-analyze-detection-kpi.mdreferences/tao-analyze-gaps-od-map.md
Showing 24 of 55 files. Browse the folder on GitHub.
Suede Skill Check: 92/100 · A
Report snapshot · · graded-v2
The rubric totals 100 points; published scores are capped at 99. Security scan: not scanned.
Source revision: 4fc9cf6cec78. Report version 84b62e4e748c. Badge URLs follow the latest listing; save the report to keep this version.
Download report · Request a fresh review
What to review next
- Description length: 3 of 5 points. 5 for 150–600 characters; 3 for 100–900; 1 for 50–1,024.
Add declarations and bundled files when they describe the skill’s actual needs. Update the source, then request a fresh review with the changed commit or PR.
tao-run-deft-object-detection passes the required frontmatter check and all seven graded checks, which cover the SKILL.md format, a declared license and a commit in the last year. Those checks earn 56 of the score’s points; the rest come from description quality, instruction depth, bundled files, maintenance and curated-list citations. How the score works.
See the checks and points
| Check | Result |
|---|---|
| Frontmatter parses and declares a name and a description | Pass |
| Name uses lowercase letters, digits and hyphens, 64 characters or fewer | Pass |
| Name matches the skill's folder name | Pass |
| Description is between 50 and 1,024 characters | Pass |
| Description says when an agent should use the skill | Pass |
| SKILL.md body is 500 lines or fewer | Pass |
| A license is declared in the frontmatter or the repository | Pass |
| Repository has a commit in the last 12 months | Pass |
| Score component | Points |
|---|---|
| Structural checks | 56 of 56 |
| Description length | 3 of 5 |
| Boundary clause | 4 of 4 |
| Example request | 3 of 3 |
| Instruction depth | 7 of 7 |
| Reference files | 3 of 3 |
| Bundled scripts | 2 of 2 |
| Recent maintenance | 9 of 9 |
| Curated-list citations | 1 of 7 |
| Tool permissions declared | 2 of 2 |
| Compatibility declared | 2 of 2 |
Two badges for tao-run-deft-object-detection
Choose the numeric badge, the letter badge, or both. Each links to this skill’s Suede listing; either valid linked badge can qualify for the feature offer.
[](https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/)
[](https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/)<a href="https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/"><img src="https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/score.svg" alt="Suede Skill Check: 92/100" width="160" height="175"></a>
<a href="https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/"><img src="https://skills.suedeai.ai/directory/nvidia/skills/tao-run-deft-object-detection/badge.svg" alt="Suede Skill Check: Grade A" width="160" height="175"></a>Suede Certified on the shield refers to this published structural quality check. Security scan: not scanned. Read the methodology.
Limited-time feature offer: display either the numeric or letter badge in your repository or skill README (both also qualify), then request badge verification. Eligible skills scoring 80–99 can receive a featured placement after Suede verifies at least one valid badge linked to the skill’s listing. Suede has not announced an end date.
Where this skill is cited
NVIDIA/skills appears in 1 curated list: VoltAgent/awesome-agent-skills.