xiao24bei/xiaobei-skill

What it solves

This project provides a collection of AI-powered "skills" designed to automate the creation of academic presentations and the conversion of static images into editable Office documents. It specifically addresses the difficulty of recreating complex academic diagrams, technical roadmaps, and research papers as editable PowerPoint slides without manually redrawing every element.

How it works

The project consists of several specialized tools that integrate with Codex to manipulate Office software:

• Image to VBA: Analyzes images (PNG/JPG) of academic diagrams or UI screenshots and generates VBA code to reconstruct them as editable Office shapes (lines, arrows, text boxes) rather than static images.
• Rebuild Image in PowerPoint: Directly operates PowerPoint to rebuild a reference image object-by-object, using a hybrid approach that combines native shapes with small raster crops for complex elements like 3D renders or microscopy images.
• Academic Paper to PPT: Transforms research papers or reports into complete, multi-page presentation decks (PPTX) suitable for thesis defenses or project reports.

Who it’s for

• Master's and PhD students
• Academic researchers
• Thesis writers
• Users who need to frequently convert static research visuals into editable presentation materials.

Highlights

• Editable-First Approach: Prioritizes the use of Office Shapes over static images to ensure the final output is fully customizable.
• Hybrid Fidelity: Retains complex visual assets (like logos or photos) as raster crops while rebuilding the structural layout with vectors.
• Manifest-Driven Reconstruction: Uses an element list and coordinate model to reduce errors in positioning and connectivity.
• Render-Verify Loop: Supports comparing exported renders with the original image to ensure accuracy when local Office software is available.
• WPS Compatibility: Includes fallback options for users of WPS Office.