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 doca-public-knowledge-mapgit 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/doca-public-knowledge-map)<a href="https://agentmods.dev/skills/nvidia/skills/doca-public-knowledge-map"><img src="https://agentmods.dev/badge/skills/nvidia/skills/doca-public-knowledge-map.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00235 | $0.03079 |
| Opus 5 | $0.00118 | $0.01540 |
| Sonnet 5 | $0.00047 | $0.00616 |
| Haiku 4.5 | $0.00023 | $0.00308 |
Grade A, and why
doca-public-knowledge-map 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 — 186 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DOCA Public Knowledge Map
Where to start: Reach for this skill whenever the question is "where does
the authoritative answer live?" — a docs page, the on-disk install layout, a
sample, or an NGC catalog entry. Read ## Public documentation entry points
first; then jump to the routing-table section that matches the user's intent.
When to load this skill
Load this skill whenever the user asks anything about NVIDIA DOCA where the
agent needs to locate authoritative information without access to the
DOCA source tree. That includes documentation-routing questions about
installing DOCA, building a sample, learning a DOCA library (Flow, DPA, Comm
Channel, GPUNetIO, …), debugging an error, finding an API or error reference,
finding a sample, release notes, or the developer forum. This skill locates
the authoritative answer; hands-on installation, building, tutorials, and
debugging route to the workflow-owning skills named in the frontmatter and
## Related skills.
This skill is intentionally a routing table, not a tutorial. Pick the entry that matches the user's intent, fetch the URL or inspect the local install path, and only then answer.
First-contact discovery — the four questions to ask before any drill-down
When a user opens with an open-ended orientation question ("I'm new with DOCA, how do I start?", "can you guide me?", "what's the easiest way to try DOCA?"), the agent does not have enough information yet to pick a path. Asking these four questions before drilling avoids wasted recommendations that the user cannot actually execute on their setup. Ask them as a single short message; do not interrogate one-at-a-time.
| Question | Why it matters | What it routes |
|---|---|---|
| 1. What OS are you on? macOS, Windows, Linux laptop, cloud VM, lab Linux box, BlueField OS itself? | DOCA installs natively only on supported Linux distributions; macOS / Windows users cannot install it at all. | Picks between the four DOCA acquisition paths in doca-setup ## no-install. macOS / Windows / no-Linux → Path 0 (NGC container). Supported Linux → Path A or B per the Installation Guide. |
| 2. What hardware do you have? No NVIDIA hardware, ConnectX SmartNIC, BlueField as a SmartNIC in a host, BlueField as a standalone DPU, not sure? | Real-traffic runtime needs a real NIC; build / read / learn does not. The user's hardware decides which DOCA libraries are even relevant. | Picks the runtime story (container is build-only without hardware). Filters which libraries make sense to learn (Flow needs a real port to do anything visible; Comch needs a host ↔ DPU pair). |
| 3. What's your goal? Just exploring, building a small first app on a specific library (Flow / RDMA / Comch / Telemetry / GPUNetIO / DPA / …), running an existing reference application, operating a service (DMS / DTS / BlueMan / Firefly), or something else? | The bundle's first-app workflow (doca-programming-guide ## modify) starts from a shipped C sample and edits down. The right sample depends on the library the user is targeting. |
Picks which library skill (if any) to load next. If the user does not yet know which library — that itself is a routing answer (see the Library- and module-specific guides table above and let the user pick). |
| 4. Which language do you plan to write the program in? C / C++, Rust, Go, Python, other? | DOCA's public surface is a C ABI. Non-C consumers go through FFI / language bindings (doca-programming-guide CAPABILITIES.md ## Capabilities and modes and the per-library skill). The C samples are the reference even when the user's language is not C. |
Picks whether the agent's first-app guidance is direct C build or FFI / bindings against the C ABI. Does not change which sample the agent points at first. |
What ships with it
5 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.
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 First seen · 186 lines · 235 tokens per session scan A 96c97dd4299a
doca-public-knowledge-map is a skill published in the GitHub repository NVIDIA/skills (3,223 stars, last pushed today), licensed Apache-2.0. It adds 235 tokens to every session and 3,079 once invoked, about $0.0012 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
spark-environment-setup
Set up a working ML training/inference environment on NVIDIA DGX Spark (GB10, aarch64, CUDA 13). Use when installing PyTorch/Unsloth/TRL/vLLM on DGX Spark, hitting libcudart or wheel-ABI errors on aarch64, or choosing between NGC containers and bare pip installs.
spark-training-gotchas
Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.
terraform-module-library
Build reusable Terraform modules for AWS, Azure, GCP, and OCI infrastructure following infrastructure-as-code best practices. Use when creating infrastructure modules, standardizing cloud provisioning, or implementing reusable IaC components.
hyperpod-version-checker
Check and compare software component versions on SageMaker HyperPod cluster nodes - NVIDIA drivers, CUDA toolkit, cuDNN, NCCL, EFA, AWS OFI NCCL, GDRCopy, MPI, Neuron SDK (Trainium/Inferentia), Python, and PyTorch. Use when checking component versions, verifying CUDA/driver compatibility, detecting version mismatches…
technical-troubleshooting
Provide setup, troubleshooting, and maintenance guidance. Use when the user reports a device that won't power on, connectivity issues, setup questions, overheating, or maintenance concerns.
gpu-optimizer
GPU optimization for consumer NVIDIA GPUs (8-24GB VRAM) covering mixed precision, gradient checkpointing, XGBoost GPU, CuPy/cuDF migration, and torch.compile. Triggers on: "optimize GPU training", "speed up CUDA", "reduce OOM", "migrate NumPy to CuPy", "manage GPU memory", "benchmark PyTorch".