Create beautiful, shareable HTML-based slide presentations styled to Red Hat brand standards. Generates single-file, self-contained HTML decks with click/keyboard navigation, story-arc-driven narrative structure, and cinematic dark-mode aesthetics. Supports embedded videos (YouTube, Vimeo, direct URLs) and linked media (memes, GIFs, images) with brand-consistent styling. Use this skill whenever the user wants to create a slide deck, presentation, quick deck, quick slides, pitch deck, talk, or briefing that should follow Red Hat branding. Also trigger when the user mentions "quick deck", "quick slides", "HTML slides", "shareable deck", "presentation for [audience]", "talk about [topic]", or asks to present technical content in Red Hat style. This skill is specifically for HTML output — if the user explicitly asks for .pptx, use the pptx skill instead, but suggest this skill as an alternative for easy sharing.
التثبيت
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
ملف SKILL.md هذا كبير جدا، لذلك يعرض SkillsMP القسم الاول فقط هنا.عرض على GitHub
name
red-hat-quick-deck
description
Create beautiful, shareable HTML-based slide presentations styled to Red Hat brand standards. Generates single-file, self-contained HTML decks with click/keyboard navigation, story-arc-driven narrative structure, and cinematic dark-mode aesthetics. Supports embedded videos (YouTube, Vimeo, direct URLs) and linked media (memes, GIFs, images) with brand-consistent styling. Use this skill whenever the user wants to create a slide deck, presentation, quick deck, quick slides, pitch deck, talk, or briefing that should follow Red Hat branding. Also trigger when the user mentions "quick deck", "quick slides", "HTML slides", "shareable deck", "presentation for [audience]", "talk about [topic]", or asks to present technical content in Red Hat style. This skill is specifically for HTML output — if the user explicitly asks for .pptx, use the pptx skill instead, but suggest this skill as an alternative for easy sharing.
Red Hat Quick Deck
Generate stunning, self-contained HTML presentations that follow Red Hat brand standards and use
narrative story arcs to make technical content compelling and memorable.
Before You Begin
Read the brand reference: references/redhat-brand.md — contains the full color palette, typography, and design rules.
Read the story arc guide: references/story-arcs.md — contains narrative structures and slide design principles.
If the redhat-brand skill is also installed, consult it for additional brand application guidance.
Ask the user which mode they prefer for the deck before generating. Use the ask_user_input tool:
Question: "What visual mode should this deck use?"
Options: "Dark mode (cinematic, dark backgrounds — best for screens & presenting)", "Light mode (clean, white backgrounds — best for print & email sharing)"
Default to Dark mode if the user doesn't specify or says "either" or "don't care."
Ask the user about PDF export intent — this determines what the printed PDF looks like and is baked into the deck's @media print CSS, so it must be decided before generation:
Question: "If/when this deck is exported as a PDF, how will it primarily be used?"
Options: "Digital sharing (email, Slack, viewing on screens — keep the on-screen palette in the PDF)", "Print (physical paper or ink-efficient PDF — force a white background with dark text)"
Default to Digital sharing if the user doesn't specify.
This only changes behavior for Dark and Expressive Dark decks. Light decks already print on white regardless of intent.
For Dark decks + Digital sharing → omit the Print Palette Override block. The PDF preserves the dark cinematic palette.
For Dark decks + Print → include the Print Palette Override block. The PDF remaps to white background with dark text and preserved Red Hat red accents.
What This Skill Produces
A single .html file that:
Is completely self-contained (inline CSS, inline JS; Google Fonts, Red Hat Design Tokens, and optionally Red Hat Icons loaded via CDN)
Can be opened in any browser, emailed, or hosted on any web server
Has keyboard navigation (arrow keys, spacebar) and click navigation
Has a slide counter / progress indicator
Looks cinematic and professional on screen, projector, or shared link
Follows a deliberate story arc that builds a persuasive narrative
Includes a contextual notes panel (toggled with 'N' key) for additional references, links, and deeper context on each slide
Supports embedded videos (YouTube, Vimeo, direct URLs) with auto-pause on slide change
Supports linked media (memes, GIFs, images) with responsive display and optional captions
Is responsive and works on mobile for async viewing
Design System
Red Hat Design Tokens
This skill integrates the official Red Hat Design Tokens (@rhds/tokens) for spacing, typography sizing,
borders, and shadows. The tokens CSS is loaded via jsDelivr CDN alongside the existing Google Fonts and
@rhds/icons dependencies. All token references include hardcoded fallbacks so decks degrade gracefully offline.
Important color note: The RHDS v3 tokens use accessibility-adjusted color values (e.g.,
--rh-color-brand-red-on-dark resolves to #FF442B, not #ee0000). This skill intentionally keeps
the established Red Hat brand palette hex values (#ee0000 for red-50, etc.) for visual consistency
with the classic brand aesthetic. Token names are annotated in comments for cross-reference.
What we use from @rhds/tokens:
Spacing: --rh-space-xs (4px) through --rh-space-7xl (128px) — consistent 4px-base scale
Typography sizing: --rh-font-size-heading-* and --rh-font-size-body-text-*
Font families: --rh-font-family-heading, --rh-font-family-body-text, --rh-font-family-code
Choose ONE color collection per deck from the Red Hat brand palette. The default and recommended
collection is "Core Dark" which matches the reference screenshot aesthetic.
Each mode sets color-scheme on :root so that any light-dark() token values resolve correctly.
Core Dark (Default)
:root { color-scheme: dark; }
--bg-primary: #000000; /* black · --rh-color-surface-darkest */--bg-secondary: #1f1f1f; /* gray-90 · --rh-color-surface-darker */--bg-surface: #292929; /* gray-80 · --rh-color-surface-dark */--text-primary: #ffffff; /* white · --rh-color-text-primary-on-dark */--text-secondary: #c7c7c7; /* gray-30 · --rh-color-text-secondary-on-dark */--text-muted: #a3a3a3; /* gray-40 */--accent: #ee0000; /* red-50 — Red Hat Red */--accent-dark: #a60000; /* red-60 */--accent-light: #f56e6e; /* red-40 */--tag-border: #383838; /* gray-70 · --rh-color-border-subtle-on-dark */--icon-filter: brightness(0) invert(1); /* white icons on dark bg */--icon-filter-accent: invert(12%) sepia(100%) saturate(10000%) hue-rotate(0deg) brightness(95%); /* red-50 icons */
Other available collections (use when the user requests a different feel):
Core Light (clean, professional — best for print/email/documentation)
:root { color-scheme: light; }
--bg-primary: #ffffff; /* white · --rh-color-surface-lightest */--bg-secondary: #f2f2f2; /* gray-10 · --rh-color-surface-lighter */--bg-surface: #e0e0e0; /* gray-20 · --rh-color-surface-light */--text-primary: #151515; /* gray-95 · --rh-color-text-primary-on-light */--text-secondary: #4d4d4d; /* gray-60 · --rh-color-text-secondary-on-light */--text-muted: #707070; /* gray-50 */--accent: #ee0000; /* red-50 — Red Hat Red */--accent-dark: #a60000; /* red-60 */--accent-light: #f56e6e; /* red-40 */--tag-border: #c7c7c7; /* gray-30 · --rh-color-border-subtle-on-light */--icon-filter: none; /* dark icons on light bg (SVGs are dark by default) */--icon-filter-accent: invert(12%) sepia(100%) saturate(10000%) hue-rotate(0deg) brightness(95%); /* red-50 icons */
When using Core Light:
The ambient glow effect uses a very subtle red tint: rgba(238,0,0,0.04)
Body and slide backgrounds alternate between --bg-primary and --bg-secondary for rhythm
Use the standard (black wordmark) logo SVG
Tag pill borders use --tag-border (#c7c7c7) with --text-secondary text
The .accent tag variant uses --accent border and text color (same as dark mode)
Architecture diagram boxes use background: rgba(242,242,242,0.8) and border: 1px solid #c7c7c7
Quote marks use --accent at low opacity
The progress indicator .current number is still red-50
Expressive Dark (for more colorful content — adds teal and purple accents)
:root { color-scheme: dark; }
--bg-primary: #1b0d33; /* purple-80 */--bg-secondary: #000000; /* black */--bg-surface: #21134d; /* purple-70 */--text-primary: #ffffff; /* white · --rh-color-text-primary-on-dark */--text-secondary: #d0c5f4; /* purple-20 */--text-muted: #b6a6e9; /* purple-30 */--accent: #ee0000; /* red-50 — Red Hat Red */--accent-dark: #a60000; /* red-60 */--accent-light: #f56e6e; /* red-40 */--highlight-teal: #37a3a3; /* teal-50 */--highlight-purple: #876fd4; /* purple-40 */--tag-border: #3d2785; /* purple-60 */--icon-filter: brightness(0) invert(1); /* white icons on dark bg */--icon-filter-accent: invert(12%) sepia(100%) saturate(10000%) hue-rotate(0deg) brightness(95%); /* red-50 icons */
Typography
@import url('https://fonts.googleapis.com/css2?family=Red+Hat+Display:wght@400;500;700;900&family=Red+Hat+Text:wght@400;500;700&family=Red+Hat+Mono:wght@400;700&display=swap');
--font-display: var(--rh-font-family-heading, 'Red Hat Display', sans-serif); /* Headlines */--font-body: var(--rh-font-family-body-text, 'Red Hat Text', sans-serif); /* Body text */--font-mono: var(--rh-font-family-code, 'Red Hat Mono', monospace); /* Code, tags, technical */
Typography Sizing (from @rhds/tokens)
Use these token-based sizes for consistent typographic scale across decks:
This creates the distinctive "red atmosphere" effect. Vary the position and opacity per slide for visual
rhythm. On some slides, shift it to the left. On the title slide, make it more prominent.
Red Hat Logo
Every deck must include the official Red Hat logo. Since these are self-contained HTML files, the logo
is embedded as an inline SVG. The skill provides two versions — use the correct one for the deck's mode.
The logo appears in two places:
Breadcrumb area (top-left of title slide) — small, subtle, as part of the navigation breadcrumb
On dark backgrounds: hat is red, wordmark is white (reverse full-color)
On light backgrounds: hat is red, wordmark is black (standard full-color)
Always maintain minimum spacing around the logo
Never distort, recolor the hat band, or add effects
Logo SVG — Reverse (for Dark Mode)
Use this on dark backgrounds. The hat is red (#e00), the wordmark is white (#fff).
This is the official RedHat-Logo-A-Reverse from static.redhat.com/libs/redhat/brand-assets/2/corp/logo--on-dark.svg.
Use this on light backgrounds. The hat is red (#e00), the wordmark is near-black (#151515).
This is the official logo extracted from the Red Hat brand standards page header.
Closing / CTA slide: Place the logo centered or left-aligned near the bottom of the slide,
above or beside the attribution, at .rh-logo.large size.
Every other slide: The logo should NOT appear on every content slide — this avoids clutter.
Instead, include a minimal red accent bar or the Red Hat red (#ee0000) in the progress indicator
to maintain brand presence throughout.
Red Hat Icons
Red Hat publishes an official icon library (@rhds/icons) with 1,135 SVGs across 4 sets. Use <img> tags
pointing to the jsDelivr CDN to add icons to slides. Icons are controlled via CSS filter variables
so they adapt to each color mode automatically.
Before using any icon, always read references/rhds-icons.md — it is the single source of truth for:
The full inventory of all 1,135 icons across 4 sets (standard, ui, microns, social)
Common alias mappings — many intuitive names don't match the actual RHDS names (e.g., database → data, integration → interoperability, build → circuit, network → network-automation)
Semantic groupings by topic (Cloud, Security, AI, DevOps, etc.) to quickly find the right icon for a slide's subject matter
Do not guess icon names. If you use a name that doesn't exist in the inventory, the icon will silently fail to load.
<!-- Basic icon --><imgclass="rh-icon"src="https://cdn.jsdelivr.net/npm/@rhds/icons@2.1.0/standard/cloud.svg"alt="Cloud"><!-- Small accent-colored icon beside a headline --><h2><imgclass="rh-icon small accent"src="https://cdn.jsdelivr.net/npm/@rhds/icons@2.1.0/standard/shield.svg"alt=""> Security First</h2><!-- Large icon for a stat slide --><imgclass="rh-icon xl accent"src="https://cdn.jsdelivr.net/npm/@rhds/icons@2.1.0/standard/graph-line-up.svg"alt=""><divclass="big-number">3.2x</div>
Icon Usage Guidelines
Use sparingly: 2-3 icons per slide maximum. Icons should clarify, not decorate.
Best uses: Stat slide topic icons, feature list bullets, architecture diagram box labels, comparison column headers.
Don't: Use icons as the sole content, mix too many sizes on one slide, use icons without supporting text.
Sizing: Use .small (24px) inline with text, .medium (48px) for feature lists, .large/.xl for hero/stat accent.
Color: Icons inherit the mode's filter by default. Use .accent class sparingly for emphasis (same restraint as red-50 text).
Spacing Scale (from @rhds/tokens)
Use the official Red Hat spacing tokens for all padding, margins, and gaps. This ensures visual
consistency with the broader Red Hat design system. All values are multiples of 4px.
The reference uses outlined pills for categorization (e.g., "Local-First", "Air-Gap Ready"). Style them
using design tokens for sizing, spacing, and borders:
<!DOCTYPE html><htmllang="en"><head><metacharset="UTF-8"><metaname="viewport"content="width=device-width, initial-scale=1.0"><title>[Deck Title]</title><!-- Red Hat Design Tokens (spacing, typography, borders, shadows) --><linkhref="https://cdn.jsdelivr.net/npm/@rhds/tokens@3.0.2/css/global.min.css"rel="stylesheet"><!-- Google Fonts --><linkrel="preconnect"href="https://fonts.googleapis.com"><linkrel="preconnect"href="https://fonts.gstatic.com"crossorigin><linkhref="https://fonts.googleapis.com/css2?family=Red+Hat+Display:wght@400;500;700;900&family=Red+Hat+Text:wght@400;500;700&family=Red+Hat+Mono:wght@400;700&display=swap"rel="stylesheet"><style>/* === RESET & BASE === */:root { color-scheme: dark; } /* Set to 'light' for Core Light mode *//* === COLOR VARIABLES (dark or light, based on user choice) === *//* === TYPOGRAPHY (uses --rh-font-family-* and --rh-font-size-* tokens with fallbacks) === *//* === SPACING (uses --rh-space-* tokens with fallbacks) === *//* === LOGO STYLES === *//* === ICON STYLES === *//* === SLIDE CONTAINER === *//* === NAVIGATION === *//* === SLIDE TYPES === *//* === CONTEXTUAL NOTES PANEL === *//* === ANIMATIONS === *//* === PRINT / PDF EXPORT === *//* @media print — see "Print / PDF Export Styles" *//* === PRINT PALETTE OVERRIDE === *//* dark-mode decks only — remaps to ink-efficient light */</style></head><body><divclass="deck"><!-- SLIDE 1: TITLE (includes logo in breadcrumb) --><divclass="slide"><divclass="breadcrumb"><!-- ⚠ REQUIRED: inline the FULL official Red Hat logo SVG from the "Red Hat Logo" section above
(lines ~220–246). Use the Reverse variant (white wordmark) for dark decks, Standard (dark
wordmark) for light decks. DO NOT substitute a stylized approximation, silhouette-only
icon, text-based wordmark, or any other placeholder — use the canonical 613×145 paths. --><svgclass="rh-logo small"xmlns="http://www.w3.org/2000/svg"viewBox="0 0 613 145"role="img"aria-label="Red Hat"><title>Red Hat</title><!-- [INLINE the 3 paths from the Reverse OR Standard logo block above, selected by deck mode] --></svg><span>› [Team / Group]</span></div><!-- ... rest of title slide ... --></div><!-- CONTENT SLIDES (no logo — brand presence via red accents) --><!-- VIDEO SLIDES use .video-container with iframe or video tag --><!-- MEDIA SLIDES use .media-container with img tag (memes, GIFs, linked images) --><!-- FINAL SLIDE: CTA (includes larger logo) --><divclass="slide"><!-- ... content ... --><divclass="logo-footer"><!-- ⚠ REQUIRED: same official SVG as above, but with class="rh-logo large" --><svgclass="rh-logo large"xmlns="http://www.w3.org/2000/svg"viewBox="0 0 613 145"role="img"aria-label="Red Hat"><title>Red Hat</title><!-- [INLINE the 3 paths from the Reverse OR Standard logo block above, selected by deck mode] --></svg></div></div></div><divclass="controls"><divclass="nav-hint">← → or click to navigate · N for additional context</div><divclass="progress"><spanclass="current">1</span> / <spanclass="total">10</span></div></div><script>// Navigation logic</script></body></html>
Required Slide Types
Every deck should include these slide types (adapt as needed):
1. Title Slide
Breadcrumb / source label (e.g., "RED HAT › NAPS STP") in small monospace, top-left
Working group or team label + date in red monospace, left-aligned
Large bold headline in Red Hat Display Black weight
Accent word(s) in the headline colored red-50
Subtitle / description in lighter text below
Tag pills for key concepts
Author attribution at bottom: "Name · Role · Organization"
2. Content Slide
Slide headline as an assertion (not a label)
Body content: paragraphs, bullet points (use sparingly), or key-value pairs
Optional: a small icon (.rh-icon.small) beside the headline for topical accent
Optional: icons as feature list markers instead of bullet characters
Optional: source attribution at bottom
3. Big Number / Stat Slide
Optional: topic icon (.rh-icon.xl.accent) centered above the big number
One large number (80-120px font size) in red-50 or white
Brief context line below in gray-30
Source attribution at bottom
4. Comparison / Before-After Slide
Two-column layout
Optional: icons representing each side as column headers (e.g., padlock-unlocked vs padlock-locked)
Clear visual distinction between old/new or with/without
Use red-50 to highlight the preferred side
5. Architecture / Diagram Slide
CSS-based box diagrams with flexbox/grid (no images required)
Boxes with borders and labels — use icons (.rh-icon.small) inside boxes alongside text labels for visual clarity
Arrows represented with CSS or Unicode characters (→, ↓)
Red-50 highlight on the key innovation
6. Quote Slide
Large pull quote in Red Hat Display, medium weight
Attribution below
Red-50 opening quotation mark as decorative element
7. Call-to-Action / Closing Slide
Clear next steps
Contact info or resources
QR code placeholder if relevant (note in contextual notes)
8. Video Slide
Use when the user provides a video URL or asks to embed a video
Supports YouTube, Vimeo (via thumbnail + click-to-embed iframe), and direct video URLs (via <video> tag)
Headline with optional accent word above the video
16:9 responsive container with brand-consistent framing (shadow, rounded corners)
Optional caption below the video for context
Source attribution at bottom
Click the play button to start the video; controls are visible for pause/seek
⚠ IMPORTANT — When a user adds a video slide, you MUST inform them:
Inline video playback requires serving this deck via HTTP or HTTPS.
YouTube and Vimeo embeds do not work when opening the HTML file directly
from your filesystem (file:// protocol) — this is a browser security
restriction that cannot be worked around.
To enable inline video playback:
Quick local server: Run python3 -m http.server in the deck's folder,
then open http://localhost:8000/deck-name.html
Or use:npx serve, VS Code Live Server, or any static file host
Or host it: Upload to any web server, GitHub Pages, S3, etc.
When opened as a local file, clicking a video will open it on YouTube in a
new browser tab instead.
This message should be delivered every time a video slide is added to a deck. Do not skip it.
YouTube iframes have multiple failure modes: file:// protocol blocks them (Error 153), ad blockers
block YouTube's tracking requests (ERR_BLOCKED_BY_CLIENT), and CSP policies can prevent embedding.
To ensure decks work reliably everywhere, always use the thumbnail + new-tab pattern:
Show the YouTube thumbnail as an image with a branded red play button overlay
On click, open the video on youtube.com in a new browser tab
The presenter's slide stays visible — they can return to it after watching
Required attribute for every .video-container: data-video-url="[full URL]"
In addition to data-video-id (used by the click-to-play handler), every video container must
carry data-video-url set to the canonical user-facing URL (e.g. https://www.youtube.com/watch?v=VIDEO_ID
or the direct .mp4 URL). The print CSS reads this attribute via content: attr(data-video-url)
to render the URL as a caption beneath the poster/thumbnail in PDF export — videos can't play in
a PDF, so readers need the link.
This approach has zero dependencies on iframes, works from any protocol, and is immune to ad blockers.
The thumbnail URL pattern: https://i.ytimg.com/vi/VIDEO_ID/maxresdefault.jpg
(falls back to hqdefault.jpg if maxres isn't available)
Required JavaScript (inside the navigation IIFE, BEFORE the navigation click handler):
// Click-to-play for video thumbnails// HTTP/HTTPS: embed iframe inline for seamless playback// file://: open in new tab (YouTube blocks iframe embeds from file:// protocol)document.addEventListener('click', (e) => {
const vc = e.target.closest('.video-container[data-video-id]');
if (vc) {
e.stopImmediatePropagation(); // prevent navigation handler from firingconst vid = vc.dataset.videoId;
if (window.location.protocol === 'file:') {
window.open('https://www.youtube.com/watch?v=' + vid, '_blank');
} else {
const iframe = document.createElement('iframe');
iframe.src = 'https://www.youtube-nocookie.com/embed/' + vid + '?autoplay=1';
iframe.setAttribute('frameborder', '0');
iframe.setAttribute('allow', 'accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture');
iframe.setAttribute('allowfullscreen', '');
vc.innerHTML = '';
vc.appendChild(iframe);
}
return;
}
});
Why e.stopImmediatePropagation(): When the click handler replaces the video container's
innerHTML, the clicked element (play button) becomes detached from the DOM. Without stopping
propagation, the navigation click handler can't find .video-container via closest() on the
detached target and accidentally advances the slide.
Inline video playback note: Include in the first slide's contextual notes and the nav hint
that videos play inline when served via HTTP (python3 -m http.server). When opened as a local
file, videos open on YouTube in a new tab instead.
<!-- YouTube video (thumbnail + new-tab pattern) --><divclass="slide"data-notes="[notes]"><divclass="slide-label animate-in">—— LABEL</div><h2class="animate-in">Headline <spanclass="accent">with accent</span></h2><divclass="video-container animate-in"data-video-id="VIDEO_ID"data-video-url="https://www.youtube.com/watch?v=VIDEO_ID"><imgclass="video-thumbnail"src="https://i.ytimg.com/vi/VIDEO_ID/maxresdefault.jpg"alt="Video description"><divclass="video-play-btn"></div></div><divclass="media-caption animate-in">Optional context about the video</div><divclass="source animate-in">Source attribution</div></div><!-- Direct video file (works from file:// natively) --><divclass="slide"data-notes="[notes]"><divclass="slide-label animate-in">—— LABEL</div><h2class="animate-in">Headline <spanclass="accent">with accent</span></h2><divclass="video-container animate-in"data-video-url="https://example.com/video.mp4"><videocontrolsmutedpreload="metadata"><sourcesrc="https://example.com/video.mp4"type="video/mp4"></video></div><divclass="media-caption animate-in">Optional context</div><divclass="source animate-in">Source attribution</div></div>
9. Media Slide (Memes, GIFs, Linked Images, Giphy)
Use when the user provides an image/meme/GIF URL or asks to include visual media
Supports Giphy GIFs (see URL conversion rules below)
Responsive container that preserves the original aspect ratio
Optional caption below for humor, context, or commentary — use a lighter tone for memes
Source / credit line at bottom
Works for memes, reaction GIFs, diagrams, screenshots, photos — any linked image
Important: Use full-resolution image URLs, not thumbnails. For imgflip templates, use the
https://imgflip.com/s/meme/Template-Name.jpg pattern (full-size) instead of
https://i.imgflip.com/2/xxxxx.jpg (150x150 thumbnails)
Giphy short https://gph.is/SHORTCODE → resolve to full URL, then extract ID for https://media.giphy.com/media/ID/giphy.gif
Giphy direct links (media.giphy.com, media0-4.giphy.com) → use as-is
Imgflip templates → use https://imgflip.com/s/meme/Template-Name.jpg (full-size)
Direct image URLs (.jpg, .png, .gif, .webp, .svg) → use as-is
Local files (via @file, file path, or "use this image") → embed as base64 data URI (see below)
Local File / @file Image Support
When the user provides a local file path or references an image via @file, embed the image
directly into the HTML as a base64 data URI. This keeps the deck fully self-contained and
portable — the image travels with the HTML file.
This works for any image the user can reference — screenshots, diagrams, downloaded memes,
photos, exported charts, etc.
Base64 increases file size by ~33%, so very large images (>5MB) may make the HTML file unwieldy.
For large images, suggest the user resize or compress first.
Animated GIFs can be embedded as base64 but will be large. For animated content, prefer Giphy
links when possible.
The user may provide images by:
Using @file syntax: @screenshot.png or @/Users/name/Desktop/diagram.png
Pasting a file path: /tmp/my-meme.jpg
Saying "use this image" with a file reference in context
Dragging an image into the conversation
<!-- Standard image/meme --><divclass="slide"data-notes="[notes]"><divclass="slide-label animate-in">—— LABEL</div><h2class="animate-in">Headline <spanclass="accent">with accent</span></h2><divclass="media-container animate-in"><imgsrc="https://example.com/meme.gif"alt="Descriptive alt text"class="slide-media"></div><divclass="media-caption animate-in">Optional witty caption or context</div><divclass="source animate-in">Source / credit</div></div><!-- Giphy GIF example --><divclass="slide"data-notes="[notes]"><divclass="slide-label animate-in">—— LABEL</div><h2class="animate-in">Headline <spanclass="accent">with accent</span></h2><divclass="media-container animate-in"><imgsrc="https://media.giphy.com/media/GIPHY_ID/giphy.gif"alt="Descriptive alt text"class="slide-media"></div><divclass="media-caption animate-in">Caption</div><divclass="source animate-in">via Giphy</div></div>
10. Thank You Slide (Required — always the final slide)
Large "Thank You" headline in Red Hat Display, Black weight
Accent word ("You") in red-50
Author name, role, and organization centered below
Red Hat logo (large) centered beneath attribution
Optionally include contact email, social handle, or team URL in small muted text
Keep it clean — no tags, no body text, generous whitespace
The ambient glow effect should be prominent on this slide for a strong visual close
Building the Narrative
When the user gives you a topic, follow this process:
Step 1: Research (if web search available)
Search for current statistics, quotes, and developments related to the topic. Look for:
Recent industry reports or surveys with quotable numbers
Expert quotes that support the thesis
Competitive landscape data
Adoption metrics or trends
Step 2: Choose a Story Arc
The arc is a spine, not a fill-in-the-blank template — vary structure so decks don't
feel formulaic. Read references/story-arcs.md and select the best arc for the content:
Problem → Tension → Resolution: Best for introducing a new tool, approach, or technology
Myth-Busting: Best for challenging conventional thinking
Journey: Best for case studies or retrospectives
Step 3: Outline the Deck
Write the slide headlines FIRST. The headlines alone should tell the complete story. Show the user
the outline before generating the full HTML if the topic is complex. Run each headline through the
anti-slop checks in references/anti-ai-slop.md before proceeding — assertions, active voice,
concrete, no hype, no em-dashes.
Step 4: Write Each Slide
For each slide:
Write the headline as an assertion
Write concise supporting content (fewer words = more impact)
Apply references/anti-ai-slop.md as you write — no buzzwords, no binary-contrast
clichés, no padded triads, no em-dashes; name the specific thing
Choose the appropriate slide type
Add source attributions where data is cited
Add contextual notes with references, links, deeper explanations, and related resources that let viewers dive deeper into the slide's topic
Step 5: Note Video & Media Opportunities
As you build the narrative, identify slides where a video or meme could strengthen the story —
a demo video after an architecture slide, a reaction meme to break tension after a dense section,
etc. Do not insert Video or Media slides during initial generation. Instead, note opportunities
in the contextual notes like:
[VIDEO OPPORTUNITY] A short demo of [feature] here would reinforce the architecture on the previous slide.
[MEME OPPORTUNITY] A well-placed meme here could break the tension after the dense comparison slide.
After delivering the deck, the post-generation prompt (see "Post-Generation: Video & Media Placement")
will guide the user to add these if they choose.
Step 6: AI Image Opportunities
As you build slides, identify moments where a custom AI-generated image would elevate the deck.
For each opportunity, add a note in the contextual notes like:
[IMAGE OPPORTUNITY] Prompt for Nano Banana Pro 2:
"[detailed image generation prompt describing the desired visual,
style, composition, colors, and mood — incorporating Red Hat brand
colors where appropriate]"
Good candidates for AI images:
Title slide hero visuals (abstract, on-brand)
Concept illustrations (e.g., "air gap" as a visual metaphor)
Background textures or atmospheric elements
Infographic-style data visualizations
Metaphorical illustrations for complex concepts
When writing prompts, specify:
Dark background compatible (for Core Dark decks)
Red Hat color palette (reds, dark grays, subtle teals/purples)
No text in the image (text will be overlaid in HTML)
Aspect ratio suited for the slide layout (usually 16:9 or specific region)
Anti-AI-Slop (Non-Negotiable)
Decks must not read as AI-written. The failure mode is not just buzzwords — it is
rhetorical sameness: padded triads, binary-contrast clichés ("not just X, it's Y"),
throat-clearing, hype adjectives, and metronomic rhythm. The full catalog, the voice
target, and the scoring rubric live in references/anti-ai-slop.md. Read it before
writing copy.
Mandatory pre-delivery gate. Before you emit the final HTML, run this gate and do
not skip it:
Quick Checks — pass every headline, eyebrow/tag, bullet, stat caption, and
contextual-notes paragraph against the banned phrases and structures in
references/anti-ai-slop.md. Rewrite every hit.
Score — rate the deck copy on the five rubric dimensions (Directness, Rhythm,
Trust, Authenticity, Density). If the total is below 35/50, revise and re-score.
Lint — run the deterministic backstop on the generated files:
node scripts/slop-lint.mjs <deck>.html <deck>.md
Fix every [error]. Review each [warning] and fix unless it is a deliberate,
defensible choice (e.g. one intentional tricolon on the closing slide).
Only then deliver. The linter catches lexical tells; the structural/rhetorical
slop is yours to catch in steps 1–2. A clean linter run is necessary, not sufficient.
Navigation JavaScript
Include this navigation system in every deck:
(function() {
const slides = document.querySelectorAll('.slide');
const currentEl = document.querySelector('.current');
const totalEl = document.querySelector('.total');
const notesPanel = document.querySelector('.notes-panel');
let idx = 0;
let notesVisible = false;
totalEl.textContent = slides.length;
functionpauseAllMedia() {
// Pause HTML5 video elements on inactive slidesdocument.querySelectorAll('.slide:not(.active) video').forEach(v => v.pause());
// Pause YouTube/Vimeo iframes on inactive slidesdocument.querySelectorAll('.slide:not(.active) iframe').forEach(f => {
f.contentWindow.postMessage('{"event":"command","func":"pauseVideo","args":""}', '*');
f.contentWindow.postMessage('{"method":"pause"}', '*');
});
}
functionplayActiveMedia() {
// Autoplay HTML5 video on the active slidedocument.querySelectorAll('.slide.active video').forEach(v => v.play());
// Autoplay YouTube/Vimeo iframes on the active slidedocument.querySelectorAll('.slide.active iframe').forEach(f => {
f.contentWindow.postMessage('{"event":"command","func":"playVideo","args":""}', '*');
f.contentWindow.postMessage('{"method":"play"}', '*');
});
}
functionshow(i) {
slides.forEach((s, j) => {
s.classList.toggle('active', j === i);
s.style.display = j === i ? 'flex' : 'none';
});
currentEl.textContent = i + 1;
pauseAllMedia();
playActiveMedia();
// Update contextual notes if visibleif (notesVisible && notesPanel) {
const note = slides[i].dataset.notes || '';
notesPanel.innerHTML = note;
}
}
functionnext() { if (idx < slides.length - 1) { idx++; show(idx); } }
functionprev() { if (idx > 0) { idx--; show(idx); } }
document.addEventListener('keydown', (e) => {
if (e.key === 'ArrowRight' || e.key === ' ') { e.preventDefault(); next(); }
if (e.key === 'ArrowLeft') { prev(); }
if (e.key === 'n' || e.key === 'N') {
notesVisible = !notesVisible;
if (notesPanel) notesPanel.classList.toggle('visible', notesVisible);
show(idx);
}
});
document.addEventListener('click', (e) => {
if (e.target.closest('.controls') || e.target.closest('.notes-panel') || e.target.closest('.video-container') || e.target.closest('.media-container')) return;
const x = e.clientX / window.innerWidth;
x > 0.5 ? next() : prev();
});
// Touch supportlet touchStartX = 0;
document.addEventListener('touchstart', (e) => { touchStartX = e.touches[0].clientX; });
document.addEventListener('touchend', (e) => {
const diff = e.changedTouches[0].clientX - touchStartX;
if (Math.abs(diff) > 50) { diff < 0 ? next() : prev(); }
});
// Print / PDF export — clear JS-set inline display so all slides render in print,// auto-fit any slide whose content would overflow the 7.5in print page, and// restore the presenter view when the dialog closes. The fit pass measures each// slide against print geometry and scales only the slides that need it; slides// that already fit print at exact 1:1 fidelity. See "Print / PDF Export Styles"// for the matching @media print CSS.window.addEventListener('beforeprint', () => {
slides.forEach(s => { s.style.display = ''; });
document.body.classList.add('printing');
slides.forEach(s => {
// Wrap children once (idempotent across repeat prints). The wrapper is a flex// container that MIRRORS the slide's own flex layout (justify-content,// align-items, text-align, gap) so vertical/horizontal centering and the// .thank-you-slide centering rule survive the wrap. Without this mirror,// the slide's `justify-content: center` is silently lost and content// top-anchors — which collides with absolutely-positioned eyebrow/breadcrumb// elements on the title slide and breaks the closer slide's centering.let fit = s.querySelector(':scope > .slide-fit');
if (!fit) {
fit = document.createElement('div');
fit.className = 'slide-fit';
while (s.firstChild) fit.appendChild(s.firstChild);
s.appendChild(fit);
}
const cs = window.getComputedStyle(s);
fit.style.cssText =
'display:flex;width:100%;height:100%;' +
'flex-direction:' + (cs.flexDirection || 'column') + ';' +
'justify-content:' + (cs.justifyContent || 'center') + ';' +
'align-items:' + (cs.alignItems || 'stretch') + ';' +
'text-align:' + (cs.textAlign || 'left') + ';' +
'gap:' + (cs.gap || '0') + ';' +
'transform-origin:top left;transform:none;';
// Lock children against flex-shrink so scrollHeight reflects true content// size — otherwise items silently shrink and clip their own content (e.g.// a code-block losing its last line of YAML, with no auto-fit triggered).for (let i = 0; i < fit.children.length; i++) {
fit.children[i].style.flexShrink = '0';
}
// Pin slide to print geometry inline so clientWidth/Height reflect the print// page even before the browser fully applies @media print rules.
s._priorStyle = s.style.cssText;
s.style.cssText +=
';display:flex !important;width:13.33in !important;height:7.5in !important;' +
'padding:0.55in 0.85in !important;box-sizing:border-box !important;' +
'overflow:hidden !important;';
// Measure content vs available area; scale down only if it overflows.const cw = fit.clientWidth, ch = fit.clientHeight;
const sw = fit.scrollWidth, sh = fit.scrollHeight;
if (sw > cw + 1 || sh > ch + 1) {
const k = Math.min(cw / sw, ch / sh);
fit.style.width = (cw / k) + 'px';
fit.style.height = (ch / k) + 'px';
fit.style.transform = 'scale(' + k + ')';
fit.dataset.fitScale = k.toFixed(3);
}
});
});
window.addEventListener('afterprint', () => {
document.body.classList.remove('printing');
// Unwrap auto-fit containers and restore inline styles so on-screen// navigation/animation behavior is unchanged.document.querySelectorAll('.slide > .slide-fit').forEach(fit => {
const s = fit.parentElement;
while (fit.firstChild) {
const child = fit.firstChild;
// Clear the flex-shrink we set during beforeprint so screen layout// returns to default flex behavior.if (child.style && child.style.flexShrink) child.style.flexShrink = '';
s.appendChild(child);
}
fit.remove();
if (s._priorStyle !== undefined) {
s.style.cssText = s._priorStyle;
delete s._priorStyle;
}
});
show(idx);
});
show(0);
})();
Animations
Use subtle entrance animations for slide content. Stagger child elements for a cinematic reveal:
Every generated deck must include an @media print block so users can produce a clean,
multi-page PDF via the browser's built-in Cmd/Ctrl+P → Save as PDF. Without these styles,
printing captures only the first slide — because the navigation JS sets inline
style="display: none" on every non-active slide, which overrides regular stylesheet rules.
The fidelity requirement (non-negotiable)
A printed PDF page must look like a shrunken on-screen slide — same headline weight, same
bullet alignment, same card geometry, same rhythm. The print CSS is not a redesign — it
only changes what's physically required for paper (ink palette, page breaks, hidden chrome,
disabled animations) and keeps everything else identical.
Failure mode to avoid: bullets that line up at a 24px offset on screen but drift to a
different offset in the PDF because em-based positioning compounded with a different font
size computed from clamp() in print context. Fix: pin print typography and geometry to
absolute pt values, and give each printed page a 16:9 aspect ratio so proportions match
the screen slide.
Why this is required
The show() function in the navigation IIFE sets s.style.display = 'none' on inactive slides.
Print CSS must use !important AND the beforeprint listener must clear the inline styles
(both are needed — some browsers honor inline display:none even in print context).
On-screen layouts use 100vh / viewport units that clip when rendered onto paper.
Entrance animations (fadeSlideUp) are triggered by the .active class and would leave
non-active slides faded/offset in print if not reset.
Navigation chrome (.controls, .notes-panel, .nav-hint) is useless in a PDF.
clamp() with vw resolves unpredictably in print context across browsers — always
override to fixed pt values for every typographic class used in the deck.
A 7.5in print page is a hard ceiling — content taller than that gets clipped silently
(the screenshot shows the title chopped at the top and the last layer chopped at the
bottom of the same slide, because the slide centers content vertically and overflows
in both directions). The navigation IIFE includes a beforeprint auto-fit pass that
wraps each slide's children in a .slide-fit div, measures it against the print page,
and applies transform: scale() only if the content overflows. Slides that already fit
print at exact 1:1 fidelity. This is a safety net — it does not replace the authoring
density budgets below.
Required @media print block (include in every deck)
/* Print preview hygiene — outside any @media block so it applies in screen context.
While the print dialog is open, the JS beforeprint handler clears each slide's
inline display:none and sets display:flex on every slide so the auto-fit
measurement pass works. In the visible viewport behind the dialog this paints
all 14 (or N) slides on top of each other (they share `position:absolute;
inset:0;`). The PDF render is unaffected — but the screen view looks like a
jumbled overlap. Hiding the deck while body.printing is set keeps the visible
viewport clean; the @media print rule below restores it for the actual page render. */body.printing.deck { visibility: hidden; }
@media print {
/* Restore visibility for the actual print render. body.printing is still set
during the print pass, but here we're in print context (not screen), and the
!important re-show overrides the screen-only rule above. */body.printing.deck,
body.printing.slide { visibility: visible !important; }
/* Custom 16:9 page — each printed page has the same aspect ratio as an on-screen slide,
so proportions are preserved. Chrome's "Save as PDF" respects this; users who pick a
physical page size in the dialog get scaled-to-fit output the browser handles. */@page { size: 13.33in7.5in; margin: 0; }
/* Preserve brand colors (red accents, logo) through print */
* { -webkit-print-color-adjust: exact; print-color-adjust: exact; }
/* Reset JS-set inline display:none and viewport-fill sizing */html, body {
height: auto !important;
width: auto !important;
overflow: visible !important;
background: #fff!important;
}
.deck {
height: auto !important; width: auto !important;
position: static !important; overflow: visible !important;
}
/* Each slide fills one 16:9 page. Dimensions and padding stay in physical units so
layout is deterministic across browsers — no viewport math in print context. */.slide {
display: flex !important;
position: relative !important;
inset: auto !important;
width: 13.33in!important;
height: 7.5in!important;
padding: 0.55in0.85in!important;
gap: 0.14in!important;
box-sizing: border-box !important;
break-after: page;
page-break-after: always;
break-inside: avoid;
page-break-inside: avoid;
overflow: hidden !important;
}
.slide:last-child { break-after: auto; page-break-after: auto; }
.slide::before, .slide::after { display: none !important; }
/* Pin typography to fixed pt values — never let clamp()/vw recompute in print.
These numbers are scaled to fit the 13.33×7.5in page at roughly the same visual
proportions as a 1920×1080 on-screen slide. Tune if your deck uses a different
screen type scale, but keep them ABSOLUTE (pt), never relative (vw/em for clamp). */.headline-xl { font-size: 60pt!important; line-height: 1.05!important; letter-spacing: -0.02em; }
.headline-lg { font-size: 42pt!important; line-height: 1.05!important; letter-spacing: -0.02em; }
.headline-md { font-size: 28pt!important; line-height: 1.1!important; }
.subtitle, .body-text { font-size: 14pt!important; line-height: 1.45!important; }
.big-number { font-size: 150pt!important; line-height: 0.9!important; letter-spacing: -0.03em; }
.quote { font-size: 30pt!important; line-height: 1.2!important; }
.slide-label, .breadcrumb { font-size: 10pt!important; letter-spacing: 0.1em; }
.attribution, .source, .nav-hint,
.arch-layer.arch-tag, .compare-col.lead, .proof-card.proof-title,
.quote-attribution, .tag { font-size: 9pt!important; }
/* Bullet geometry: pin ALL dimensions in pt so marks align identically to screen. */.bullet-list { gap: 10pt!important; max-width: none !important; }
.bullet-listli {
padding-left: 18pt!important;
font-size: 12pt!important;
line-height: 1.45!important;
}
.bullet-listli::before {
content: ''!important;
position: absolute !important;
left: 0!important;
top: 0.7em!important;
width: 10pt!important;
height: 2pt!important;
background: #ee0000!important;
border-radius: 0!important;
}
/* Cards/grids: preserve on-screen structure; just normalize padding to physical units. */.compare-grid, .proof-grid { gap: 0.18in!important; }
.compare-col, .proof-card { padding: 0.2in!important; }
/* Stack-layer slides pack a lot of vertical content (5 layers + 4 arrows + headline +
body) — give them tighter padding/leading so the auto-fit scaler isn't forced to
shrink the type below ~80% to make room. */.arch-layer { padding: 0.12in0.2in!important; }
.arch-stack { gap: 2pt!important; margin-top: 0.1in!important; }
.arch-layer.arch-tag { margin-bottom: 0!important; }
.arch-layerh4 { font-size: 11pt!important; line-height: 1.15!important; margin-bottom: 2pt!important; }
.arch-layerp { font-size: 9pt!important; line-height: 1.3!important; }
.arch-arrow { font-size: 11pt!important; line-height: 1!important; }
/* Code blocks: in print, force overflow:visible. The on-screen rule uses
overflow:hidden (NEVER overflow-x:auto — per CSS spec, an `auto` on one axis
forces the other to `auto` as well, and the resulting overflow-y:auto silently
clips the bottom YAML line on tall code samples). Tighter line-height fits more
lines per slide before the auto-fit scaler kicks in. */.code-block {
font-size: 9pt!important;
line-height: 1.45!important;
padding: 0.16in!important;
overflow: visible !important;
}
/* Breadcrumb and attribution stay absolutely positioned (same as screen) so the
title slide's geometry is preserved. Anchor to the 16:9 print canvas. */.breadcrumb { top: 0.4in!important; left: 0.85in!important; }
.attribution { bottom: 0.4in!important; left: 0.85in!important; }
/* Disable entrance animations — render final state */.slide.animate-in,
.slide.active.animate-in {
animation: none !important;
opacity: 1!important;
transform: none !important;
}
/* Hide presenter chrome */.controls, .nav-hint, .notes-panel { display: none !important; }
/* Ambient glow / decorative backgrounds waste ink on paper */.ambient-glow, .bg-glow, [class*="glow"] { display: none !important; }
/* Video slides: keep thumbnail, drop play button + iframe, show URL caption */.video-container.video-play-btn,
.video-container.play-button,
.video-containeriframe { display: none !important; }
.video-container { aspect-ratio: auto; max-height: 4.5in; }
.video-containerimg { max-height: 4.5in; object-fit: contain; }
.video-container[data-video-url]::after {
content: "▶ "attr(data-video-url);
display: block;
font-family: var(--rh-font-family-code, 'Red Hat Mono', monospace);
font-size: 10pt;
color: #6a6e73;
margin-top: 6pt;
word-break: break-all;
}
/* Media slides: keep images on-page */.media-container { max-height: 5in!important; page-break-inside: avoid; }
.media-containerimg, .slide-media { max-height: 5in; object-fit: contain; }
}
Print fit budget — what fits on a 7.5in page
The auto-fit JS in the navigation IIFE will scale any slide that overflows so nothing
clips, but a slide that has to scale below ~80% looks visibly smaller than its neighbors
in the PDF — the typography reads as off-rhythm, even though no content is lost. Author
slides to fit 1:1, and split when in doubt. Use these per-pattern budgets:
Stacked architecture layers (.arch-stack) — max 4 layer cards if the slide
also has a title + intro paragraph. 5+ layers → split into two slides (e.g.,
"Layers 1–3" / "Layers 4–5") or drop the intro paragraph and use a tighter
one-line lead-in.
Bullet list — max 6 bullets at default 12pt body. 7+ → split, or condense to
one-line bullets.
Compare grid (.compare-col), proof grid (.proof-grid) — max 4 items per
column when the slide also has a title + intro paragraph.
Code blocks — max 18 lines of YAML/code at default font. Longer → trim to the
relevant section (use ... for elided fields), or use a continuation slide for the
remainder.
Big-number slide (.big-number) — max 2 numbers per slide. 3+ → split.
Body paragraph — max ~280 characters (≈ 3 lines at print width) above stacked
cards or grids; up to ~450 characters if no other content sits below.
When a slide has both a title + intro + dense visual element (stack, grid, code, or
multiple big numbers), prefer splitting over compressing. Two clean slides read better
in the PDF than one auto-shrunk slide.
Layout pitfalls — three traps that look fine on screen but break in PDF
These three CSS patterns will pass on-screen review and silently break the PDF export.
Watch for them when authoring or editing slide CSS:
Stat-row align-items: baseline with a giant numeral. A 130pt+ digit paired with
a 9–14pt descriptor on align-items: baseline puts the descriptor at the typographic
baseline of the digit — i.e., visually pinned to the bottom edge of the numeral, with
~80% of the numeral's vertical area as empty space above. Use align-items: center
for any stat row where the figure and descriptor differ in size by more than ~3×.
.code-block { overflow-x: auto; }. Per the CSS spec, setting one overflow axis
to auto forces the other axis to auto too. The result: a tall YAML sample with
~18+ lines will silently get its bottom line clipped at the box edge in print, even
though the content is visible on screen (where scrollbars solve it). Use
overflow: hidden for the on-screen rule and overflow: visible !important in
@media print. The .slide already has overflow: hidden, which clips any
horizontal long-line escapes at the slide boundary instead.
Auto-fit .slide-fit wrapper that doesn't mirror the slide's flex layout. The
beforeprint handler wraps each slide's children so they can be measured and
transform-scaled. If the wrapper uses a generic display:flex; flex-direction:column;
without copying the slide's justify-content / align-items / text-align /
gap, the slide's justify-content: center and the .thank-you-slide
centering rules silently disappear — content top-anchors, the title slide's
slide-label collides with the absolutely-positioned breadcrumb, and the closer
slide's "Thank You" sits at the top of an otherwise empty page. Always read the
slide's computed style and mirror its flex properties onto the wrapper. Also set
flex-shrink: 0 on every wrapped child — otherwise items silently shrink, the
scrollHeight measurement matches clientHeight, the auto-fit scaler doesn't
trigger, and individual elements (especially .code-block) clip their own content.
No print-preview hygiene — visible viewport jumbles all slides during print dialog.
When the user opens the print dialog, the JS beforeprint handler clears each
slide's inline display:none and sets display:flex on every slide so the auto-fit
measurement pass works. In the visible viewport behind the dialog, this paints
all N slides on top of each other at the same position:absolute; inset:0;
coordinates — text from every slide visibly overlaps. The PDF preview pane and the
final PDF are correct; only the dimmed page behind the dialog looks like a jumbled
mess. Add this rule outside any @media block: body.printing .deck { visibility: hidden; }, plus a matching @media print { body.printing .deck, body.printing .slide { visibility: visible !important; } } to restore the deck for the actual
print render. Without these, every print dialog opens to a chaotic-looking page
behind it — users may think the deck is broken.
Print Palette Override that forgets h4, stat descriptors, or the closer name.
For Dark+Print decks, the override remaps .text-primary (white) elements to dark
and .text-secondary (light gray) elements to muted dark. The trap: it's easy to
list h1, h2, h3 and miss h4 (used by .arch-layer headers in stack slides),
.compare-col h3 (card titles, technically caught by the h3 selector but worth
listing explicitly), .thank-you-slide .author-block .name (the presenter's name on
the closer), .big-number-context (stat descriptors), and .arch-layer p (stack
body text). When any of these is forgotten, the affected text renders white-on-white
in the printed PDF — invisible to the reader, undetectable on a screen review.
Audit the override every time you add a new component class that uses
var(--text-primary) or var(--text-secondary).
This block is conditional — include it only when both of these are true:
The deck uses Core Dark or Expressive Dark mode, AND
The user said the PDF export is for Print (see Step 5 of "Before You Begin")
When included, it remaps the printed output to an ink-efficient white-on-dark-text palette
while preserving Red Hat red accents.
Skip this block when:
The deck is Core Light (on-screen palette already prints well), OR
The deck is Dark and the user said the PDF is for Digital sharing — in that case
the PDF should preserve the cinematic dark palette so it matches the on-screen experience
when viewed in email, Slack, or a PDF reader on a screen.
@media print {
:root { color-scheme: light; }
body, .deck, .slide { background: #fff!important; color: #151515!important; }
/* ⚠ Every element with on-screen `color: var(--text-primary)` (white) MUST be listed
here, otherwise it renders white-on-white in print. Easy to miss: h4 (used by
.arch-layer headers), .compare-col h3, .thank-you-slide .author-block .name. If
you add a new component that renders headline-style text, add its selector here. */.headline, h1, h2, h3, h4,
.headline-xl, .headline-lg, .headline-md,
.big-number, .quote, .stat-number,
.arch-layerh4,
.compare-colh3, .proof-cardh4,
.thank-you-slide.author-block.name { color: #151515!important; }
/* ⚠ Same trap one shade lighter — anything that uses `var(--text-secondary)` (light
gray) on screen renders as nearly-invisible light gray on white in print unless
remapped here. Includes .big-number-context (stat descriptors), .arch-layer p
(stack body text), and .thank-you-slide .author-block (closer subtitle). */.subtitle, .body-text, p, li, .bullet-listli, .media-caption,
.compare-colp, .proof-cardp,
.arch-layerp,
.big-number-context,
.thank-you-slide.author-block,
.thank-you-slide.author-blockdiv { color: #3c3f42!important; }
/* ⚠ Bold labels inside bullets/cards — on-screen they're `var(--text-primary)` (white);
in print they must be dark. Forgetting this rule causes `<strong>` text to render
white-on-white — invisible but still occupying layout width, which looks like a
giant gap between the bullet dash and the visible text. Must override explicitly. */.bullet-listlistrong, .compare-colstrong, .proof-cardstrong,
.big-number-contextstrong,
pstrong, listrong { color: #151515!important; font-weight: 700!important; }
.breadcrumb, .slide-label, .nav-label, .source, .attribution,
.quote-attribution, .tag,
.arch-layer.arch-tag, .compare-col.lead, .proof-card.proof-title {
color: #6a6e73!important;
}
/* Preserve Red Hat red-50 on accents */.accent, .red-accent, [class*="red-50"],
.big-number, .compare-col.winner.lead,
.proof-card.proof-title { color: #ee0000!important; }
/* Tag pills: dark outline on white */.tag, [class*="tag"] {
background: transparent !important;
border-color: #3c3f42!important;
color: #151515!important;
}
.tag.hot, .compare-col.winner { border-color: #ee0000!important; }
.compare-col, .proof-card, .arch-layer {
background: #fff!important;
border-color: #c7c7c7!important;
}
.arch-layer.os-added {
border-color: #ee0000!important;
background: rgba(238, 0, 0, 0.04) !important;
}
.proof-card { border-left-color: #ee0000!important; }
/* Swap reverse (white) wordmark → dark for print. Targets the wordmark path by its
original fill — the canonical 613×145 logo has the wordmark as the last path with
fill="#fff". Fedora silhouette (fill="#e00") stays red in both modes per brand. */body.printing.rh-logopath[fill="#fff"] { fill: #151515!important; }
}
If the deck uses a light mode, or if the deck is dark but the user chose Digital sharing
as the PDF intent, omit the palette override block entirely (but still include the base
@media print block above — slide pagination, hidden chrome, and pinned typography are
required regardless of palette).
For dark decks with Digital-sharing intent, you also need to drop the background: #fff !important;
declarations on html, body and .deck from the base print block above so the dark backdrop
carries through to the PDF. Replace those #fff values with var(--bg-primary) (or the
equivalent dark hex) so the PDF reads as a faithful screen capture.
Before delivering the HTML file, verify:
User was asked about dark or light mode preference before generation
User was asked about PDF export intent (digital sharing vs print) before generation, and the print CSS reflects their choice
Print fit verified — opened the deck in a browser, triggered Cmd/Ctrl+P → Save as PDF, and visually scanned the preview pane for any slide where the headline, last bullet, last card, or last code line is clipped. The auto-fit JS will scale overflow slides to fit; if any slide scales below ~80% (look for typography that reads visibly smaller than neighbors), the content is over-budget and should be split per "Print fit budget"
All text uses Red Hat font family (Display, Text, or Mono)
Red-50 (#ee0000) appears on every slide (even if just in the nav or a small accent)
Red Hat logo appears on the title slide (breadcrumb area, small) and closing slide (larger)
Logo uses the correct variant: reverse (white wordmark) for dark, standard (black wordmark) for light
Logo is the official inline SVG from the "Red Hat Logo" section above (viewBox 0 0 613 145, all three canonical paths) — never a stylized approximation, silhouette-only icon, text-in-SVG wordmark, or placeholder
Logo is inline SVG (no external image dependencies)
Color palette matches the chosen mode (Core Dark or Core Light or Expressive Dark)
Color contrast meets WCAG AA (4.5:1 for body text, 3:1 for large headlines)
Headlines tell a complete story when read in sequence
Keyboard navigation works (← → Space N)
Click/tap navigation works
Contextual notes are present with references, links, and additional context for deeper exploration
Sources are attributed on data slides
At least one AI image opportunity is noted in contextual notes
Icons (if used) load from jsDelivr CDN and display correctly in chosen mode
Icons are used sparingly (2-3 per slide max) and enhance rather than clutter
Red Hat Design Tokens CSS is loaded via jsDelivr CDN (@rhds/tokens@3.0.2/css/global.min.css)
color-scheme is set correctly on :root (dark for Core Dark / Expressive Dark, light for Core Light)
Spacing uses --rh-space-* tokens with px fallbacks (e.g., var(--rh-space-lg, 24px))
Typography sizing uses --rh-font-size-* tokens with fallbacks where appropriate
Tags use --rh-border-radius-pill and --rh-space-* for padding
Video slides (if any) use the thumbnail + click-to-embed pattern with data-video-id attributes
Video slides (if any) include the protocol-aware JS handler (inline iframe on HTTP, new tab on file://)
User was informed that inline video requires serving via HTTP/HTTPS (mandatory notification)
Media slides (if any) have descriptive alt text on <img> tags
Media/video containers are excluded from click-to-navigate (click guard in JS)
File is self-contained (no external dependencies besides Google Fonts, jsDelivr icons, jsDelivr design tokens, and user-provided video/image URLs)
Progress indicator shows current/total slides
The narrative follows a clear story arc with emotional rhythm
Thank You slide is present as the final slide with author name, role, and Red Hat logo
The accent word(s) in the title slide headline are colored red-50 for emphasis
Video/media opportunities are noted in contextual notes (not inserted during initial generation)
Post-generation prompt was shown to the user after delivering the deck, offering to add videos or memes
@media print block is present and resets inline display:none with !important so all slides render in PDF export
Print CSS uses a custom 16:9 page (@page { size: 13.33in 7.5in; margin: 0; }), fixed pt typography for every class (no clamp()/vw leaking into print), and pt-based bullet geometry so list marks align identically to screen — the PDF must look like a shrunken on-screen slide, not a reinterpreted print view
beforeprint / afterprint listeners are wired in the navigation IIFE (clear inline styles before print, restore show(idx) after)
For dark-mode decks: print palette override block is included (white background, #151515 body text, red-50 accents preserved, wordmark-path fill swapped via body.printing .rh-logo path[fill="#fff"]); for light-mode decks, this block is omitted
Print palette override remaps <strong> inside bullets and cards to #151515 — otherwise bold labels inherit the dark-mode white and render invisibly on paper, producing a large gap before the visible bullet text
Every .video-container carries a data-video-url attribute so the print CSS caption renders
A Markdown outline companion file is saved alongside the HTML (same kebab-case filename, .md extension)
Example: Mapping the OpenCode Screenshot to This System
The reference screenshot ("Your AI Assistant Should Live Where You Work") demonstrates:
Breadcrumb: [RH Logo SVG] › NAPS STP [logo small + font-mono, small, muted]
Label: —— NAPS STP WORKING GROUP · FEB 27, 2026 [font-mono, red-50, small, tracking-wide]
Headline: Your AI Assistant [Red Hat Display, Black, white, ~64px]
Should Live Where You Work ["Live Where You Work" in red-50, italic]
Subtitle: A fully local, air-gap-ready AI... [font-body, gray-40, ~18px]
Tags: [Local-First] [Air-Gap Ready] ... [pill style, monospace, outlined]
Attribution: Todd Wardzinski · Architect · Red Hat [font-body, gray-40, small]
Background: Black with subtle red radial glow [upper-right corner, very low opacity]
This is the target aesthetic. Every title slide should feel this cinematic and intentional.
File Delivery
Save two files per deck — both use the same kebab-case filename derived from the deck title:
/mnt/user-data/outputs/[deck-name].html — the interactive, self-contained HTML deck
/mnt/user-data/outputs/[deck-name].md — a Markdown outline companion for editable
leave-behinds (wiki paste, email, notes apps)
Present both files to the user. Mention that the HTML can be saved as a PDF via
Cmd/Ctrl+P → Save as PDF (landscape orientation), and that the .md file mirrors the deck
as plain text for sharing.
Markdown Companion Format
The .md file should mirror the deck's narrative in plain Markdown. Use this template:
# [Deck Title]
[Subtitle / tagline — optional]
**[Team / Group]** · [Date]
_Author Name · Role · Red Hat_
---
## Slide 1 — [Headline]
[Body copy, bullets, or stat — whatever is on the slide, rendered as Markdown]
- Bullet one
- Bullet two
> Contextual notes: [the contents of `data-notes` for this slide, including references and links]**Source:** [source attribution if present]
---
## Slide 2 — [Headline]
[...]
---
## Slide N — Thank You
Author Name · Role · Red Hat
[Contact info if present]
Guidelines for the Markdown companion:
One ## Slide N — Headline per slide, in order.
Preserve accent words as emphasis (*word* or **word**), not HTML.
