paperbanana-slide-deck

PaperBanana Slide Deck

End-to-end slide deck creation orchestrated via Nexus RDIV workflow.

Pipeline: Content -> Style Selection -> Outline -> Prompts -> PaperBanana Image Gen -> PPTX/PDF Merge

When to Use

• User wants to create a slide deck / presentation from content (paper, notes, topic)
• User says "make slides", "create a presentation", "PPT", "slide deck"
• User has content and wants it turned into visual slides

Dependencies

Component	Role	Location
PaperBanana CLI	Image generation	python -m paperbanana.cli slide-batch
Style libraries	150+ style presets (4 sources)	See R1 for discovery paths
Baoyu slide-deck assets	Merge scripts	~/.claude/plugins/cache/baoyu-skills/content-skills/*/skills/baoyu-slide-deck/
Bun	Run merge scripts	bun or npx -y bun

Nexus Integration

This skill runs as a Nexus RDIV task. Initialize via:

Route: IMPLEMENTATION
Complexity: MEDIUM
Workflow: RDIV

RDIV Workflow

Phase R: Research — Content Analysis & Style Selection

Step R1: Discover style libraries — Scan all four sources and merge into a unified list.

# Source 0 (PRIMARY): PaperBanana built-in library (123 structured styles)
SKILL_DIR="$(find ~/.claude/plugins/cache -path '*/paperbanana-slide-deck/references/styles' -type d 2>/dev/null | head -1)"

# Source 1: baoyu-slide-deck styles (16 slide-optimized)
BAOYU_SLIDE=$(ls -td ~/.claude/plugins/cache/baoyu-skills/content-skills/*/skills/baoyu-slide-deck/references/styles 2>/dev/null | head -1)

# Source 2: baoyu-infographic styles (20 visual-rich)
BAOYU_INFOG=$(ls -td ~/.claude/plugins/cache/baoyu-skills/content-skills/*/skills/baoyu-infographic/references/styles 2>/dev/null | head -1)

# Source 3: theme-factory themes (10 color/typography presets)
THEME_FACTORY=$(ls -td ~/.claude/plugins/cache/anthropic-agent-skills/document-skills/*/skills/theme-factory/themes 2>/dev/null | head -1)

Merge: Glob all found directories for *.md, deduplicate by filename (priority: baoyu-slide-deck > built-in > baoyu-infographic > theme-factory). Present the full unified list to user during R3.

If none found: Use PaperBanana's 23 code-level presets (fallback).

Step R2: Analyze content

Read user's input content. Analyze for:

• Topic/discipline (biology, ML, business, etc.)
• Audience (academic, executive, general)
• Tone (formal, educational, energetic)
• Content length -> recommended slide count:

Content Length	Recommended Slides
< 1000 words	5-10
1000-3000 words	10-18
3000-5000 words	15-25
> 5000 words	20-30

Step R3: Style recommendation

Based on content signals, recommend 2-3 styles from the 150+ presets. Core mappings:

Content Signals	Recommended Style
biology, chemistry, medical	scientific
machine learning, AI, deep learning	ml-ai
biotech, genomics, drug discovery	biotech
neuroscience, brain, cognitive	neuroscience
ecology, climate, sustainability	environmental
architecture, system design	blueprint
business, investor, corporate	corporate
product launch, keynote	bold-editorial
Apple/Tesla premium	tech-keynote
finance, investment, consulting	financial-elite
tutorial, education, beginner	sketch-notes
classroom, teaching	chalkboard
academic research, briefing	intuition-machine
SaaS, product, dashboard	notion
executive, minimal	minimal
entertainment, gaming	dark-atmospheric
innovation, design showcase	creative-bold
creative, children	vector-illustration
historical, heritage	vintage
lifestyle, wellness, artistic	watercolor
science communication	editorial-infographic
storytelling, magical	fantasy-animation
gaming, retro, developer	pixel-art

Step R4: Interactive selection

Use AskUserQuestion for each decision:

Q1: Style

header: "Style"
question: "Which visual style for this deck?"
options:
  - label: "{recommended} (Recommended)"
    description: "Best match based on content analysis"
  - label: "{alternative_1}"
    description: "[description]"
  - label: "{alternative_2}"
    description: "[description]"

Q2: Audience

header: "Audience"
question: "Who is the primary reader?"
options:
  - label: "General readers"
  - label: "Beginners/learners"
  - label: "Experts/professionals"
  - label: "Executives"

Q3: Slide Count

header: "Slides"
question: "How many slides?"
options:
  - label: "{N} slides (Recommended)"
  - label: "Fewer ({N-3} slides)"
  - label: "More ({N+3} slides)"

Step R5: Save analysis

Create output directory and save analysis.md:

slide-deck/{topic-slug}/analysis.md

Phase D: Design — Outline & Prompts

Step D1: Load style spec

If baoyu available: Read ${BAOYU_DIR}/references/styles/{style_name}.md If fallback: Use paperbanana.guidelines.slide_styles.get_style_prompt(style_name)

Step D2: Generate outline

Create outline.md with structure:

# {Deck Title}

Style: {style_name}
Slides: {N}
Audience: {audience}

## Slide 1: Cover
- Layout: title-hero
- Title: {deck title}
- Subtitle: {subtitle}

## Slide 2: {title}
- Layout: {layout_type}
- Key points:
  - Point 1
  - Point 2

...

## Slide N: Back Cover
- Layout: title-hero
- Content: Key takeaway or CTA

Available layouts (from PaperBanana):

• title-hero, quote-callout, key-stat, split-screen
• icon-grid, two-columns, three-columns, image-caption, bullet-list
• linear-progression, binary-comparison, comparison-matrix
• hierarchical-layers, hub-spoke, bento-grid, funnel
• dashboard, venn-diagram, circular-flow, winding-roadmap

Step D3: Optional outline review

Ask user if they want to review the outline before proceeding.

Step D4: Generate prompts

For each slide, create prompts/NN-slide-{slug}.md:

{style_instructions from Step D1}

---

Generate a presentation slide image with the following content:

**Slide Title:** {title}
**Layout:** {layout_type}
**Content:**
{content points}

**Design Requirements:**
- Resolution: 16:9 aspect ratio (1920x1080 or 3840x2160)
- No slide numbers, page numbers, footers, headers, or logos
- All text must be clear, readable, and correctly spelled
- One clear focal point per slide

Phase I: Implement — Generate & Merge

Step I1: Batch generate images

cd <paperbanana_dir>
python -m paperbanana.cli slide-batch \
  --prompts-dir "slide-deck/{topic-slug}/prompts" \
  --output-dir "slide-deck/{topic-slug}" \
  --style {style_name} \
  --resolution 4k \
  --config configs/fast.yaml \
  --iterations 2 --auto

Step I2: Merge to PPTX

BAOYU_DIR=$(ls -td ~/.claude/plugins/cache/baoyu-skills/content-skills/*/skills/baoyu-slide-deck 2>/dev/null | head -1)
bun "${BAOYU_DIR}/scripts/merge-to-pptx.ts" "slide-deck/{topic-slug}"

Step I3: Merge to PDF

bun "${BAOYU_DIR}/scripts/merge-to-pdf.ts" "slide-deck/{topic-slug}"

If baoyu merge scripts unavailable: inform user that PNG images are ready but PPTX/PDF merge requires baoyu-slide-deck plugin.

Phase V: Verify — Review & Iterate

Step V1: Display results

Read and display each generated slide image. Report:

• Total slides generated
• Style used
• Output directory
• PPTX/PDF file paths

Step V2: User feedback

Ask user:

header: "Review"
question: "How do the slides look?"
options:
  - label: "Looks great, done!"
  - label: "Regenerate specific slides"
    description: "I'll tell you which ones to redo"
  - label: "Change style and redo all"
    description: "Start over with a different style"

Step V3: Selective regeneration

If user wants to modify specific slides:

1. User identifies slide numbers
2. User provides feedback for each
3. Update the prompt file: prompts/NN-slide-{slug}.md
4. Regenerate: python -m paperbanana.cli slide --input prompts/NN-slide-{slug}.md --style {style_name} --resolution 4k --config configs/fast.yaml --auto
5. Re-merge PPTX/PDF

Output Structure

slide-deck/{topic-slug}/
  source.md              <- Original content
  analysis.md            <- R phase output
  outline.md             <- D phase output
  prompts/               <- D phase output
    01-slide-cover.md
    02-slide-xxx.md
    ...
  01-slide-cover.png     <- I phase output
  02-slide-xxx.png
  {topic-slug}.pptx      <- I phase output
  {topic-slug}.pdf       <- I phase output

Fallback Behavior

Component Missing	Fallback
All style libraries	Use slide_styles.py 23 code-level presets
Baoyu merge scripts	Output PNG only, skip PPTX/PDF
Bun/npx	Skip PPTX/PDF merge, inform user

Quick Reference

# List available styles
python -m paperbanana.cli slide --list-styles

# Generate single slide with style
python -m paperbanana.cli slide --input prompt.md --style scientific

# Batch generate with style
python -m paperbanana.cli slide-batch --prompts-dir prompts/ --style blueprint --resolution 4k