NVIDIA/skills is a catalogue of portable instruction sets that teach coding agents how to use NVIDIA software for robotics, simulation, CUDA, retrieval-augmented generation, and related workflows. Developers install these skills in agents such as Claude Code or Codex, while the catalogue mirrors skills maintained in separate NVIDIA product repositories.
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 NVIDIA/skills --skill deepstream-devgit clone --depth 1 https://github.com/NVIDIA/skillsWrote 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/nvidia/skills/deepstream-dev)<a href="https://agentmods.dev/skills/nvidia/skills/deepstream-dev"><img src="https://agentmods.dev/badge/skills/nvidia/skills/deepstream-dev/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/nvidia/skills/deepstream-dev"><img src="https://agentmods.dev/badge/skills/nvidia/skills/deepstream-dev.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- Socket pass
- Snyk pass
- NVIDIA SkillSpector pass
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 | $0.00047 | $0.03535 |
| Opus 5.5 | $0.00019 | $0.01414 |
| Sonnet 5.5 | $0.00009 | $0.00707 |
| Haiku 4.5 | $0.00005 | $0.00353 |
Grade A, and why
deepstream-dev 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 1mo 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 — 201 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DeepStream Development Skill
This skill requires access to all of the reference documents listed in the references/ directory below. Ensure they are available before executing the workflow.
When this skill is active, ALWAYS read the relevant reference documents before generating code. Do NOT rely on memory - the reference documents contain critical details about exact property names, correct API usage, and common pitfalls.
SDK and Architecture Quick Reference
DeepStream SDK Version Requirements
- GStreamer: 1.24.2
- NVIDIA Driver: 590+
- CUDA: 13.1
- TensorRT: 10.14.1.48
- Platforms: Ubuntu 24.04 (x86_64 and ARM64/Jetson)
Typical Pipeline Flow
Source → Stream Muxer → Inference → [Tracker] → OSD → Renderer
Components in [brackets] are optional -- only add them when the user explicitly requests them.
| Stage | Role | Key Element(s) | Required? |
|---|---|---|---|
| Source | Input from files, RTSP, cameras | nvurisrcbin (preferred), nvmultiurisrcbin, filesrc |
Yes |
| Stream Muxer | Batches streams for inference | nvstreammux |
Yes |
| Inference | TensorRT model execution | nvinfer, nvinferserver |
Yes |
| Tracker | Multi-object tracking across frames | nvtracker |
Only if requested |
| OSD | Draws bounding boxes, labels, overlays | nvosdbin |
Yes (for visualization) |
| Renderer | Display or save output | nveglglessink, nv3dsink, filesink |
Yes |
Memory Model
DeepStream uses NVIDIA Video Memory Manager (NVMM) for zero-copy GPU buffer transfers. Caps strings use memory:NVMM to indicate GPU memory (e.g., video/x-raw(memory:NVMM), format=NV12).
Critical Rules
- Only Add Requested Components: Do NOT add pipeline elements the user did not ask for.
- Tracker (
nvtracker): Only add when the user explicitly requests tracking or object IDs across frames - Secondary GIEs: Only add when the user requests classification or attribute extraction
- Analytics (
nvdsanalytics): Only add when the user requests line crossing, ROI counting, etc. - Message broker (
nvmsgbroker/nvmsgconv): Only add when the user requests Kafka/cloud messaging - When in doubt, build the minimal working pipeline and let the user ask for additions
- Tracker (
What ships with it
20 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- BENCHMARK.md 4.6 KB
- evals/evals.json 12 KB
- references/best_practices.md 33 KB
- references/buffer_apis.md 54 KB
- references/docker_containers.md 9.0 KB
- references/gstreamer_plugins.md 31 KB
- references/kafka_messaging.md 59 KB
- references/media_extractor_advanced.md 27 KB
- references/metamux_config.md 12 KB
- references/nvds_msgapi_adapter.md 28 KB
- references/nvinfer_config.md 22 KB
- references/rest_api_dynamic.md 11 KB
- references/service_maker_api.md 61 KB
- references/streaming_sources.md 16 KB
- references/tracker_config.md 49 KB
- references/troubleshooting.md 25 KB
- references/use_cases_pipelines.md 35 KB
- references/utilities_config.md 45 KB
- skill-card.md 5.0 KB
- skill.oms.sig 8.3 KB
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.
- 1mo ago Changed · +18 lines 8d49cda5b235
- 1mo ago First seen · 183 lines · 47 tokens per session scan A 3e42a832ea21
deepstream-dev is a skill published in the GitHub repository NVIDIA/skills (3,622 stars, last pushed today), licensed Apache-2.0. It adds 47 tokens to every session and 3,535 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
deep-learning-pytorch
Use when writing or changing PyTorch code — model, Dataset/DataLoader, training and evaluation loops, determinism, mixed precision, torch.compile, checkpoints with safetensors, Hugging Face models, ONNX export, and tests that run on CPU.
model-serving-and-monitoring
Use when a model is served online (FastAPI, BentoML) or scored in batch, when changing how a model artifact is loaded or versioned, or when adding drift, quality or prediction monitoring and a rollout plan for a new model.
data-pipeline-orchestration
Use when adding or changing a scheduled data or ML pipeline — Dagster assets, Prefect flows, Airflow DAGs or a cron'd entrypoint — idempotent partitioned stages, retries, backfills, quality gates, and testing the pipeline without the scheduler.
data-validation-contracts
Use when data crosses a boundary into your code (file, API, warehouse table, upload, model input or output), when a join or cleaning step can change row counts, or when adding pandera, Great Expectations, Pydantic or dbt checks.
experiment-tracking-and-reproducibility
Use when a change trains a model, reports a metric, or touches seeds, data or model versions, the lockfile, MLflow or DVC — what every run must log, seeding, environment pinning, data versioning and keeping tests off the shared tracking server.
leakage-safe-feature-engineering
Use when building or changing features, preprocessing, encoders, resampling or train/validation/test splits — keep every fitted transform inside a Pipeline fitted on training folds only, and build time and group features without seeing the future or the same entity twice.