aicr-creating-slide-decks

par nvidia

À utiliser lors de la création d'un diaporama HTML autonome ou d'un support visuel pour un concept technique ou un flux de travail (par exemple un demos/*.html) — affiché en plein écran ou…

npx skills add https://github.com/nvidia/aicr --skill aicr-creating-slide-decks

Creating Slide Decks

Overview

A deck here is one self-contained HTML file — inline CSS and inline SVG, no build, no dependencies, no network — that opens in any browser and projects cleanly. One idea per slide; hand-drawn SVG diagrams carry the weight. It's the talk-to-it companion to a runnable demo script (see the aicr-creating-guided-demos skill).

Copy skeleton.html: it's already a working deck (palette + chrome + keyboard nav + a title slide + a content slide + an SVG diagram stub). Add one <section class="slide"> per idea; the script auto-counts and navigates.

When to use

  • Building a visual talking-point (a demos/*.html) to present or teach a concept/workflow, full-screen or projected.
  • A leave-behind a reader can click through on their own later.
  • Not for: prose docs (use Markdown), anything needing a real web app/build, or live data dashboards.

Conventions

  • One self-contained file. Inline <style> + inline <svg>; no external fonts/CSS/JS/CDN — it must work offline and on a projector with no network. Lives in demos/.
  • Full-viewport slides + nav. One <section class="slide" data-title="…"> per idea; only .active shows. Keyboard (←/→/Space, Home/End), on-screen ‹ ›, a progress bar, a slide counter, F fullscreen, and #n deep-links — all from the skeleton's script.
  • One idea per slide. An eyebrow label + a heading + one visual (a diagram, a code block, or a small table). If it overflows, split it.
  • Inline SVG carries the concept. Hand-author with the node/edge classes (node/node-g, edge/edge-g, nlabel/nsub/elabel); one diagram per slide; a viewBox plus max-height: …vh so it scales. Keep labels legible from the back of a room.
  • Dark NVIDIA theme. The :root palette (green #76b900 accent, NVIDIA Sans with system fallback). Wordmark in text — no fabricated NVIDIA logo.

Rules

  • Real, accurate content — no marketing fluff. The deck explains the actual thing; don't overclaim. (Same honesty bar as the demo it accompanies.)
  • Self-contained or it doesn't ship. A single external dependency breaks the offline/projector promise.
  • Commit-ready by default. Generic content, no personal/machine specifics, no invented logos.

Validate before you ship

Decks fail silently — a malformed <svg> just doesn't render. Before committing:

  • Parse every inline <svg> as XML. Inline SVG is held to a stricter bar than the surrounding HTML: a bare &, an unclosed tag, or a stray attribute that a browser tolerates in HTML makes the whole <svg> fail to draw. Escape & as &amp;, </> as &lt;/&gt;. Quick check:

    python3 - <<'PY'
    import re; from xml.dom import minidom
    html = open("demos/your-deck.html").read()
    for n, s in enumerate(re.findall(r'<svg\b.*?</svg>', html, re.DOTALL), 1):
        minidom.parseString(s); print(f"svg #{n}: well-formed")
    PY
    
  • Open it in a browser and arrow through every slide; check each diagram/code block fits the viewport (cap max-height; slides scroll internally as a fallback). If you can't open a browser, at minimum confirm each diagram's viewBox height stays within its max-height cap.

Worked example

demos/evidence-demo-slides.html — nine slides (problem → mechanics → diagrams → payoff → roadmap), four hand-built SVG diagrams, keyboard nav, dark theme.

Common mistakes

  • External fonts/CSS/JS → dies offline or on the projector. Inline everything.
  • Wall of text per slide → one idea per slide; split instead.
  • Fabricated logo or marketing fluff → text wordmark, real content.
  • SVG taller than the viewport → set max-height; verify in a browser.

Plus de skills de nvidia

fhir-basics
nvidia
Apprend aux agents comment fonctionnent les API FHIR R4, quelles ressources sont disponibles, comment les interroger avec des paramètres de recherche, et comment analyser correctement tous les formats de réponse…
compileiq-validate-result
nvidia
Utiliser APRÈS qu'une recherche soit terminée et AVANT de réclamer un accélérateur ou d'expédier un ACF. Charge le CSV dump_results, extrait les K meilleurs candidats (mono-objectif)…
changelog-audit
nvidia
Auditer le CHANGELOG.md de Warp avant une publication : récupérer les entrées perdues, trier par impact utilisateur, affiner le langage des entrées, ajuster les retours à la ligne et (en mode branche de publication) mettre à jour la comparaison…
dgx-diagnose
nvidia
Diagnostiquer les problèmes courants du DGX Station GB300 — plantages CUDA, ciblage incorrect du GPU, bugs de conteneur vLLM/SGLang, problèmes d'état MIG, erreurs NVLink/Fabric Manager,…
aicr-managing-openvex
nvidia
Use when adding, updating, or removing CVE/GHSA suppressions in `.openvex.json` — the OpenVEX document consumed by the daily image vulnerability scan workflow.…
aicr-creating-guided-demos
nvidia
Génère un script de démonstration guidée interactive (demos/*.sh), en direct ou à son rythme, avec le modèle Frame → Tell → Show → Close. Se déclenche sur « script de démonstration », « guidé…
aicr-analyzing-snapshots
nvidia
À utiliser lors de l'analyse d'un fichier YAML de snapshot AICR, de l'examen de l'état du cluster, de la comparaison des caractéristiques des fournisseurs, de l'extraction d'informations sur la topologie GPU/réseau, ou…
aicr-release-notes
nvidia
À utiliser lors de la rédaction du résumé lisible par un humain des notes de version GitHub pour une prochaine version d'AICR. Se déclenche sur « notes de version », « brouillon de version »…