Add image-based workflow for PPTX skill with CI

[robtaylor]

Summary

This PR adds an image-based workflow for the PPTX skill that handles complex HTML slides (SVG diagrams, CSS gradients, intricate layouts) that the existing html2pptx converter struggles with.

Key additions:

• Puppeteer-based HTML→PNG rendering (render-slides.js)
• PPTX creation from PNG images with diagram overlay support (create-from-images.js)
• Proper dependency management with package.json + npm and pyproject.toml + uv
• Lock files for reproducible builds
• Comprehensive test suite with Claude-style test slides
• CI workflow with GitHub Actions (3 jobs: Node.js, Python deps, Integration)

Test Slides

The test suite includes slides that simulate what Claude would generate for professional presentations:

Slide	Description
architecture-slide.html	System architecture with SVG diagrams, gradients, arrows
metrics-slide.html	Dashboard with KPI cards, line chart, donut chart
diagram-overlay.html	Standalone data flow diagram for overlay testing
test-slide.html	Basic slide with SVG and CSS gradients

Test Plan

• [x] CI passes (Node.js workflow: 24s, Python deps: 10s, Integration: 28s)
• [x] Tests render 4 complex HTML slides to PNG
• [x] Tests create PPTX with diagram overlays
• [x] Tests verify PPTX structure (ZIP signature)

Files Changed (15 files, +4122/-10)

Category	Files
Scripts	render-slides.js, create-from-images.js
Dependencies	package.json, package-lock.json, pyproject.toml, uv.lock
Documentation	SKILL.md (updated)
Tests	test-workflow.js, 4 test slide HTML files
CI	.github/workflows/test-pptx-skill.yml
Config	.gitignore (updated)

🤖 Generated with Claude Code