NVIDIA Cloud Functions is a platform for deploying, managing, and running GPU-based workloads such as inference, streaming, and batch processing, while routing requests to GPU worker clusters. Its monorepo contains the service code, deployment materials, documentation, examples, command-line tools, agent skills, and validation tools used to build and operate the platform.
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/nvcf --skill nvaf-architecture-diagramsgit clone --depth 1 https://github.com/NVIDIA/nvcfWrote 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/nvcf/nvaf-architecture-diagrams)<a href="https://agentmods.dev/skills/nvidia/nvcf/nvaf-architecture-diagrams"><img src="https://agentmods.dev/badge/skills/nvidia/nvcf/nvaf-architecture-diagrams/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/nvcf/nvaf-architecture-diagrams"><img src="https://agentmods.dev/badge/skills/nvidia/nvcf/nvaf-architecture-diagrams.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00070 | $0.01227 |
| Opus 5.5 | $0.00028 | $0.00491 |
| Sonnet 5.5 | $0.00014 | $0.00245 |
| Haiku 4.5 | $0.00007 | $0.00123 |
Grade A, and why
nvaf-architecture-diagrams 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 5d 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 — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
NVAF Architecture Diagrams
Apply the NVIDIA Cloud Architecture Diagrams Style Guide to NVAF figures while preserving the technical meaning requested by the user.
Sources of truth
Use this precedence:
- The user's explicit content and deliverable requirements.
- NVIDIA Cloud Architecture Diagrams Style Guide v1.1.
- Repository conventions in docs/AGENTS.md.
- Existing figures as implementation references.
Before editing, read the style-guide summary and the provenance file for the relevant assets. For the turn-orchestration figure, read its asset map. Treat nvaf-turn-orchestration.svg as a layout and validation example, not authority when it conflicts with the style guide.
Authoring contract
- Use this visual system only for architecture diagrams. Do not force workflow or software-stack content into it.
- Establish one clear primary path before adding secondary or optional paths.
- Use groups only when they contain multiple related components.
- Select icons by component semantics. Reuse the supplied generic icon when no specific icon exists; do not invent one.
- Keep source icons byte-identical. Do not recolor, crop, flip, rotate, or otherwise modify them.
- Pick one standard component-icon size for the diagram. Use the small size for technical diagrams and the large size only for simple executive diagrams.
- Keep labels short, normally six words or fewer. Break a long approved name onto two lines or shorten it after showing the full name once; never solve overflow by shrinking one label.
- Align components and orthogonal paths to a grid. Reserve open connector lanes between groups and rows instead of routing through cards.
- For a cross-group connector, use only a short, direct stub from the source component to its nearest group edge; put lateral routing in the open lane outside the group.
- Land each arrow on the midpoint of its target edge with a straight final segment at least 16 pt long. Avoid backtracking, U-turns, and a bend immediately before the arrowhead.
- Put a numbered callout on a straight, open part of its path—not on a bend, arrowhead, group boundary, or another connector. Keep parallel callouts at least one callout diameter apart.
- Use the guide's 1 pt stroke and 4 pt arrowhead proportions. Arrowheads must touch the target edge and point cleanly into it at normal reading size.
- Snap a group-level path to the group edge and a component-level path to the component itself.
- Use the guide's existing path styles and palette. Do not invent new path colors, icon treatments, or group styles.
- Keep the architecture canvas on a white canvas. For a dark presentation surface, place the unchanged white diagram on a white panel over the dark surround; do not invert or recolor the diagram.
- Embed local PNG icons as data-URI symbols in tracked SVGs so GitLab renders them without external file loading. Keep a
data-sourceattribute pointing to the repository asset.
What ships with it
2 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.
- 5d ago First seen · 88 lines · 70 tokens per session scan A 1316f0c78d70
nvaf-architecture-diagrams is a skill published in the GitHub repository NVIDIA/nvcf (253 stars, last pushed 3d ago), licensed Apache-2.0. It adds 70 tokens to every session and 1,227 once invoked, about $0.0003 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 2026-11-01.
nvaf-architecture-diagrams compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| nvaf-architecture-diagramsNVIDIA/nvcf | Skill | 253 | 3d ago | Apache-2.0 | A | 70 | Claude Code, Codex |
| dstack-prototypingdstackai/dstack | Skill | 2.3k | yesterday | MPL-2.0 | A | 80 | Claude Code, Codex |
| runpodericrisco/rsc-harness | Skill | 503 | 2d ago | MIT | A | 114 | Claude Code, Codex |
| tao-run-on-kubernetesNVIDIA-TAO/tao-skill-bank | Skill | 99 | yesterday | Apache-2.0 | C | 80 | Claude Code |
| vmware-privateaivmware-skills/VMware-PrivateAI | Skill | 0 | 2d ago | MIT | B | 237 | Claude Code |
| kubernetes-mesh-provisionerKnuckles-Team/container-manager-mcp | Skill | 6 | yesterday | MIT | C | 60 | Claude Code, Codex |
| kilocode-fleet-statusPOWERFULMOVES/PMOVES.AI | Skill | 7 | today | none | A | 46 | Claude Code, Codex |