norman-fig2ppt
norman-fig2ppt
Before reconstruction, ask whether to use Quick or Advanced mode and wait for the answer. If the user has already requested quick, advanced, high-fidelity, or element-by-element reconstruction, use the corresponding mode without asking again. Default to English for prompts, explanations, reports, and generated code comments, and to PowerPoint as the output host. Preserve the source image's text unless the user requests translation; honor the user's explicit language, layout, and save-location choices. If Excel or Word is explicitly requested, generate VBA for that host; the bundled runner supports PowerPoint only.
Select the engine for the current system
Derive the skill root from the absolute path of this SKILL.md; do not hardcode a username or installation location. Determine the operating system first: on Windows, use bin/windows-x64/norman-engine.exe with PowerShell's & and a quoted absolute path; on macOS, use /bin/sh "<skill-root>/bin/norman-engine". The Mac launcher automatically selects the bundled Apple Silicon or Intel runtime. Users do not need to install Python, a compiler, or MCP, and the engine does not download dependencies on first launch. Do not run the EXE on Mac, call system Python, or use Windows DLLs as Mac runtime libraries.
Windows examples:
& "<skill-root>/bin/windows-x64/norman-engine.exe" probe
& "<skill-root>/bin/windows-x64/norman-engine.exe" map <image-width-px> <image-height-px> <canvas-width-pt> <canvas-height-pt> --pretty
& "<skill-root>/bin/windows-x64/norman-engine.exe" crop "<source-image>" "<assets-directory>" --region "R01:x:y:w:h" --manifest "<crop-manifest.json>" --pretty
& "<skill-root>/bin/windows-x64/norman-engine.exe" check "<module.bas>" --mode advanced --pretty
& "<skill-root>/bin/windows-x64/norman-engine.exe" diff "<source-image>" "<rendered-image>" --mode advanced --pretty
& "<skill-root>/bin/windows-x64/norman-engine.exe" run "<module.bas>" --macro BuildFinal --output "<new-output.pptx>" --pretty