Frontend Slides Engine

Frontend Slides Engine

Use this copy as the HTML renderer and visual-discovery engine. It does not preserve native editable PowerPoint objects.

Route

• New HTML deck: load html-template.md, viewport-base.css, and animation-patterns.md.
• Style discovery: load STYLE_PRESETS.md and bold-template-pack/selection-index.json, then only the shortlisted template recipes.
• PPTX intake: run scripts/extract-pptx.py; label the result semantic extraction/redesign rather than fidelity conversion.
• PDF export: run scripts/export-pdf.mjs; use the legacy .sh exporter only on a compatible Unix shell.
• Load references/upstream-skill.md only for a detail not covered here.

Output contract

• Keep one fixed 16:9 stage and use visibility/opacity/pointer-events for slide switching.
• Produce one HTML file when single-file delivery is requested; inline local CSS, JavaScript, fonts, and images, and reject unexpected runtime network dependencies.
• Provide keyboard navigation, visible focus, reduced-motion behavior, and print styles.
• Preserve source order, text, images, and notes during PPTX semantic redesign; do not claim chart/master/animation fidelity.

QA

Open the deck in the configured browser, test first/last slide boundaries and every navigation key, inspect console errors, render all slides, and export a PDF whose page count matches the deck. Treat accessibility and offline delivery as partial until they are tested.

Recovery

• Missing Playwright/browser: deliver verified HTML and mark PDF/render checks partial.
• Broken external font: use a bundled or system fallback and recheck wrapping.
• Oversized content: shorten copy or select a higher-capacity layout before reducing type below the quality floor.