Coordinate the slide-text-layer-inpaint, slide-image-dual-render, and slide-visual-polish-qa Skills to convert source slide images into a final editable PPTX and HTML with hardlocked validation, visual QA, repair-wave planning, and final delivery packaging. Use when the user asks to run or plan an end-to-end editable deck conversion, automate repair waves, close fail/blocking slides, summarize visual QA backlog, or package final slide reconstruction artifacts.
Reconstruct slide images into both editable PPTX and standalone HTML from one backend-agnostic renderer. Use when slide screenshots or exported PNGs need faithful native-object reconstruction, crop metadata, visual QA, and package validation.
Preprocess slide source images by mapping semantic text, pseudo text, micro text, decorative glyph-like text, and unresolved text-like regions; generate text and inpainting masks; and create clean background assets for downstream faithful slide reconstruction. Use before slide-image-dual-render when source screenshots need a separated text layer and controlled background cleanup. This skill never edits PPTX/HTML renderer files and never weakens slide-image-dual-render hardlock gates.
Companion visual QA for editable slide reconstruction workflows. Use after slide-image-dual-render has produced PPTX/HTML slide outputs and work artifacts to compare source slide images against rendered PPTX rasters and HTML screenshots, compute perceptual/pixel/edge differences, classify visual fidelity issues, and generate actionable polish fix plans without generating, repairing, or directly editing PPTX/HTML.