slide-deck

Slide Deck Creator

Create stunning, interactive HTML slide deck presentations backed by deep research — academic papers, case studies, and real-world examples. Every deck is a self-contained HTML file with D3.js visualizations, Anime.js animations, keyboard/click navigation, and embedded figures extracted from referenced papers. Plus a companion markdown file with speaker notes.

Dependencies

This skill uses other skills from the claude-tricks plugin:

• alphaxiv-paper-lookup — for fetching structured overviews of arxiv papers
• pdf — for extracting figures, tables, and diagrams from paper PDFs
• frontend-design — for visual design principles and aesthetics

Arguments

• $ARGUMENTS - The topic, theme, or subject for the presentation

---

Phase 0: Requirements Clarification

Before any research or design, ask the user targeted questions to shape the presentation:

• Audience: Who is this for? (executives, engineers, students, general public, investors?)
• Goal: What should the audience walk away with? (convince, educate, inspire, inform?)
• Depth: High-level overview or deep technical dive?
• Duration: How many slides / how long is the talk? (5-min lightning talk vs 45-min keynote)
• Tone: Formal/corporate, casual/fun, academic, provocative, storytelling?
• Must-include: Any specific papers, case studies, data points, or examples to feature?
• Must-avoid: Any topics, framing, or competitors to steer clear of?
• Branding: Any color scheme, logo, or visual identity to match?
• Code: Should slides include code snippets or live demos?
• Format preferences: Dark/light theme? Any reference presentations they admire?

Present questions as a concise numbered list. WAIT for the user to answer before proceeding.

---

Phase 1: Research (Parallel Agents)

After getting requirements, launch parallel research agents to gather material.

Agent 1 — Academic Research & Figure Extraction

Research academic foundations for: $ARGUMENTS

Step 1: Find papers Use WebSearch to find relevant arxiv paper IDs on the topic.

Step 2: Get paper overviews via alphaxiv For each promising paper, fetch its structured overview:

curl -s "https://alphaxiv.org/overview/{PAPER_ID}.md"

If the overview doesn't have enough detail, fetch the full text:

curl -s "https://alphaxiv.org/abs/{PAPER_ID}.md"

For each paper, extract:

• Key concepts and contributions
• Important diagrams, figures, tables, or results worth showing in slides
• Quotable findings or conclusions
• How it connects to the presentation topic

Step 3: Extract figures and diagrams from papers For papers with important visuals (architecture diagrams, result tables, charts, figures), download the PDF and extract the images using the pdf skill:

# Download the paper PDF
curl -sL "https://arxiv.org/pdf/{PAPER_ID}" -o /tmp/{PAPER_ID}.pdf

# Convert relevant pages to images
python3 scripts/convert_pdf_to_images.py /tmp/{PAPER_ID}.pdf /tmp/{PAPER_ID}_pages/

Use pdfplumber to identify figure bounding boxes and extract specific figures:

import pdfplumber
from PIL import Image

pdf = pdfplumber.open(f"/tmp/{PAPER_ID}.pdf")
page = pdf.pages[PAGE_NUMBER]
# Extract image at bounding box coordinates
im = page.to_image(resolution=300)
im.crop((x0, y0, x1, y1)).save(f"/tmp/{PAPER_ID}_fig{N}.png")

Convert extracted images to base64 for embedding in the HTML deck:

base64 -i /tmp/{PAPER_ID}_fig{N}.png

Return: a structured summary of 3-8 relevant papers with extracted insights, ordered by relevance. For each paper with useful visuals, include the base64-encoded figure data and a caption/citation.

Also search for survey papers that provide broader context.