Create an app demonstration whose motion supports readable actions. This skill preserves the black-MacBook production method in a small runnable scaffold; it contains no proprietary app media, voice, font, or model binary.
- Inspect the existing app, films, approved pacing, colors, and available licensed assets. Preserve requested deliverables; put an additional compilation in its own directory. Read workflow.md for capture, native sharpness, and timeline decisions.
- Reuse an existing production renderer where available. For a new project, run
python3 scripts/scaffold.py DEST --model /path/to/macbook-pro.glbfrom this skill. The supported asset and attribution are in model-license.md. The generated demo is synthetic, silent, and 12 seconds long; replace it with authentic app states for a product film. - Read hardware.md before adjusting materials, UVs, camera, lid motion, particles, or UI lifts. Its exact settings prevent the recurring gray-casing, soft-display, and camera-chasing-hinge failures. The bundled Three.js modules implement these settings.
- Build the narration/action/camera timeline together. Every claimed result must exist in the captured app. Use measured CSS coordinates for clicks and highlights; actual UI crops for floating details. Read audio-and-compilation.md when adding voice, sound, or combining shorts.
- Run the scaffold smoke/preview checks before a full render. Follow validation.md on the encoded result: hardware must read black from all shown angles, app text must remain sharp, actions must match voice, and model credits must accompany delivery.
Default this style to dark backgrounds, the app's own accent colors, restrained headings, and one clear feature action per shot. Opening, a full turn, or a close-and-exit belong at transitions; hold the laptop still enough to read the app. Do not add chapter numbers, persistent decorative dividers, footer chrome, invented UI, or unrelated palette choices unless requested. User preferences override these style defaults.
Credentials stay in environment variables; cache approved narration and alignment instead of regenerating it for visual revisions. A request to create videos does not authorize publishing them or their source. Preserve model/media licensing independently from this skill's MIT-owned code.