Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add nebius/nebius-physical-ai --skill physical-ai-data-factorygit clone --depth 1 https://github.com/nebius/nebius-physical-aiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nebius/nebius-physical-ai/physical-ai-data-factory)<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/physical-ai-data-factory"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/physical-ai-data-factory/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/physical-ai-data-factory"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/physical-ai-data-factory.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 3 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium MCP Rug Pull · line 296 Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.Fix: Pin the image: image:tag or image@sha256:abc123
- medium MCP Rug Pull · line 298 Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.Fix: Pin the image: image:tag or image@sha256:abc123
- low Tool Misuse · line 298 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00078 | $0.09804 |
| Opus 5 | $0.00039 | $0.04902 |
| Sonnet 5 | $0.00016 | $0.01961 |
| Haiku 4.5 | $0.00008 | $0.00980 |
Grade A, and why
physical-ai-data-factory scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 618 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Physical AI Data Factory (NPA-native, no OSMO)
Source And Attribution
NPA-native re-implementation of the NVIDIA Physical AI Data Factory / Video Data
Augmentation workflow. Design adapted from NVIDIA agent skills
(https://github.com/NVIDIA/skills), primarily physical-ai-video-data-augmentation.
Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. Upstream licenses:
Apache-2.0 and CC-BY-4.0. See skills/NOTICE-NVIDIA-SKILLS. NPA orchestrates on
SkyPilot (not OSMO) and composes existing workbench tools.
Three NVIDIA components in the pipeline are the real open-source projects, not
NPA look-alikes: Cosmos Transfer 2.5 augments, Cosmos Evaluator
(https://github.com/nvidia-cosmos/cosmos-evaluator, Apache-2.0) grades, and
Cosmos Curator (https://github.com/nvidia-cosmos/cosmos-curate, Apache-2.0)
curates. See skills/NOTICE-NVIDIA-COSMOS-OSS for exactly which upstream code
runs and where NPA substitutes its own endpoint.
When To Use
Load this skill when the user wants to author, validate, submit, run, or view the
physical-ai-data-factory.yaml blueprint, adapt it to a new dataset, run it on
GPUs, or troubleshoot why a run's Rerun panel / augmented output looks wrong.
Do NOT invent an npa workbench data-factory tool — there is none. The blueprint
is pure composition of existing toolRefs; only add real tools with tests.
What It Is
The independent paidf-cosmos3.yaml variant is documented at
docs/workbench/guides/paidf-cosmos3.md. It uses real source-video-conditioned
Cosmos 3 video2video generation and does not replace or silently change this
skill's Cosmos Transfer 2.5 blueprint.
workflows/testing/physical-ai-data-factory.yaml — one
npa.workflow/v0.0.1 spec. Blueprint → NPA stage mapping:
| NVIDIA stage | NPA state | Tool (all REAL — no stubs) | Runtime |
|---|---|---|---|
| Config Generation | generate-configs |
data_factory_stages.generate_configs (run.shell) |
CPU |
| Understand & Annotate | annotate-original |
workbench.token_factory.caption |
Token Factory (zero-GPU) |
| Augment & Multiply | augment |
workbench.cosmos2.transfer_execute (real Cosmos Transfer 2.5 --execute; uploads video+frames to S3) |
GPU |
| Evaluate & Validate | grade loop (evaluate + quality-gate) |
workbench.cosmos_evaluator.evaluate (real Cosmos Evaluator: hallucination + attribute verification) + data_factory_stages.grade_gate |
Token Factory + CPU |
| Pseudo-Label Augmented | annotate-augmented |
npa workbench token-factory caption (run.shell) |
Token Factory |
| Curation | cosmos-curate |
workbench.cosmos_curate.curate (real Cosmos Curator stages → clips/ + metas/v0/) |
CPU |
| Curation review | curate |
workbench.fiftyone.curate_augmented (real FiftyOne Brain, fail closed, merges the curator report) |
CPU |
| Visualize | visualize (accepted) / visualize-rejected (rejected evidence) |
workbench.nurec.visualize → data_factory_viz.build_run_rrd → reports/sim2real.rrd |
CPU, prebuilt npa-rerun-viewer image |
| Finalize | finalize |
data_factory_stages.finalize (real aggregate report) |
CPU |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago Changed · +31 lines ecb5e832c346
- 5d ago Changed fb5f2dd5bf91
- 7d ago First seen · 587 lines · 78 tokens per session scan A 7c33b23af1e7
physical-ai-data-factory is a skill published in the GitHub repository nebius/nebius-physical-ai (28 stars, last pushed today), licensed Apache-2.0. It adds 78 tokens to every session and 9,804 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
tensorrt-llm
High-throughput LLM inference on NVIDIA GPUs.
google-cloud-solution-guided-gke-ai-migration
Guides the migration of existing AI workloads (Cloud Run, Gemini API, Gemini Enterprise Agent Platform) to self-hosted GKE inference using gcloud and kubectl. Use when the user has an existing AI inference workload (on Cloud Run, the Gemini API, Gemini Enterprise Agent Platform, or a custom VM) and wants to move it to…
agent-platform-tuning
Agent Platform Model Tuning. Use when you need to fine-tune open models or Gemini models using Agent Platform infrastructure. Don't use for model training outside Agent Platform, model deployment to endpoints (use agent-platform-deploy), or managing serving endpoints (use agent-platform-endpoint-management).
gke-inference
Deploys and optimizes AI/ML inference workloads on GKE, using GPUs, TPUs, and model servers. Use when deploying GKE inference servers, configuring GKE GPU resources for inference, or deploying LLMs on GKE. Don't use for generic batch jobs or HPC task queues (use gke-batch-hpc instead).
agent-platform-endpoint-management
Manages Agent Platform serving endpoints. Use when you need to create, list, describe, update, or delete serving endpoints for model deployment on Agent Platform. Also use when troubleshooting endpoint permission, quota, or resource busy errors. Don't use for deploying models to endpoints or for running model…
modal
Modal is a serverless cloud platform for running Python on demand, including on-demand GPUs. Use when deploying or serving AI/ML models, running GPU-accelerated workloads (training, fine-tuning, inference), serving web endpoints, scheduling batch jobs, or scaling Python code to cloud containers with the Modal SDK.