NVIDIA Elements is a design system and agent-oriented toolkit for building operational interfaces for AI/ML, robotics, and autonomous-vehicle software. Its CLI and MCP tools provide component APIs, design tokens, examples, imports, validation, and setup guidance across several web frameworks.
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/elements --skill guidance-webgpu-performancegit clone --depth 1 https://github.com/NVIDIA/elementsWrote 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/elements/guidance-webgpu-performance)<a href="https://agentmods.dev/skills/nvidia/elements/guidance-webgpu-performance"><img src="https://agentmods.dev/badge/skills/nvidia/elements/guidance-webgpu-performance/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/elements/guidance-webgpu-performance"><img src="https://agentmods.dev/badge/skills/nvidia/elements/guidance-webgpu-performance.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- 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 | $0.00098 | $0.01836 |
| Opus 5.5 | $0.00039 | $0.00734 |
| Sonnet 5.5 | $0.00020 | $0.00367 |
| Haiku 4.5 | $0.00010 | $0.00184 |
Grade A, and why
guidance-webgpu-performance 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 — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Optimize WebGPU Performance
Find the limiting resource before changing the renderer. Produce a reproducible diagnosis that separates JavaScript preparation, browser/API submission, data transfer, and GPU execution, then make only changes supported by measurements.
Required context
- Read the repository
AGENTS.mdand the affected project'sDEVELOPMENT.md. - Inspect the render loop, update path, buffer and texture ownership, shader code, and resource lifecycle involved in the representative workload.
- Before adding or changing a benchmark, inspect the affected package's
package.json, local Vitest benchmark configuration, existing*.test.bench.tsfiles, and any shared configuration they import. Do not assume another project's harness applies unchanged. - Read the performance diagnostic playbook when classifying a bottleneck, selecting instrumentation, or proposing an optimization.
- Read the reusable WebGPU test suite guide when adding or reviewing a
*.test.webgpu.tssuite, choosing a test mode, or deciding whether a performance task belongs in CI.
Establish the performance question
Define the workload and target before measuring:
- Separate initialization, steady rendering, interactive updates, streaming updates, picking/readback, and cleanup. Do not combine them into one unexplained metric.
- Record object and primitive counts, update frequency and size, draw/pass counts, canvas pixel dimensions, device-pixel ratio, transparency, anti-aliasing, and relevant renderer settings.
- Choose the user-visible metric: frame budget and tail latency, update latency, throughput, startup time, or memory. FPS alone can hide missed frames behind the display's refresh-rate ceiling.
- Preserve correctness and representative visual output. A faster result that omits required work is not an optimization.
- Inspect
git status --shortand preserve unrelated changes. Record when measurements include a dirty worktree.
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.
- 1mo ago Changed · +14 lines · +15 tokens per session 3f00e7939df8
- 1mo ago First seen · 89 lines · 83 tokens per session scan A d6d9f4e3a599
guidance-webgpu-performance is a skill published in the GitHub repository NVIDIA/elements (102 stars, last pushed today), licensed Apache-2.0. It adds 98 tokens to every session and 1,836 once invoked, about $0.0004 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 2026-09-03.
Other skills, from other repositories
Cypress v14 Component Testing
Component testing patterns with Cypress v14 including React, Vue, and Angular component mounting, custom mount commands, interaction testing, visual snapshots, and integration with Vite and Webpack bundlers.
angular
Use when building, refactoring, or debugging Angular (v20/21+): standalone components, signals, zoneless change detection, @if/@for/@defer control flow, inject() DI, resource()/httpResource(), RxJS interop, NgRx SignalStore, ng CLI. NOT React (that is react), NOT Next.js (that is nextjs), NOT a TypeScript language…
module-federation
Add, review, or debug client-rendered Module Federation support in Rsbuild-first TypeScript applications. Covers host/remote roles, exposes, remotes, shared dependencies, generated types, runtime loading failures, and observability checks.
vue-application-structure
Establishes or reviews the directory layout, component conventions, composable design, Pinia store structure, and Vue Router configuration for a Vue 3 TypeScript application. Invoked when the user asks to structure a Vue app, set up the project layout, or review Vue architecture.
react-application-structure
Establishes or reviews the directory layout, feature boundaries, state design, routing approach, and data-fetching conventions for a React 18+ TypeScript application. Invoked when the user asks to structure a React app, set up a scalable architecture, or review React project organization.
vue-component-design
Designs or reviews Vue 3 component APIs. Handles prop drilling decisions, applies compound component patterns with provide/inject, defines minimal public API surfaces with typed props/emits/slots, and enforces Vue 3.4+ composition conventions. Invoked when creating new components, refactoring existing ones, or…