rebuild-site

motiscope: rebuild-site

Rebuild a whole landing page from a walkthrough recording. This is analyze + recreate applied to an entire page, and it leans on the same division of labor:

• The numbers give you the WHEN + the structure — motiscope measures section timing, transitions, and per-section entrance easing, and curates the frames that show each section.
• You give the WHAT — from the frames you reconstruct the sections, layout, copy, design system, and what kind of animation each uses (open vocabulary — scroll reveals, pins, masks, draws, whatever you see).
• Imagen fills the photos — generate the photographic assets the design needs.

It's a team effort: measured timing + curated frames + your vision + generated assets → a full page. Be honest about which is which in the final summary.

Resolve scripts + preflight

SCRIPTS="${CLAUDE_PLUGIN_ROOT:-}/scripts"; REFS="${CLAUDE_PLUGIN_ROOT:-}/references"
[ -f "$SCRIPTS/ingest.py" ] || { SCRIPTS="<abs dir of this SKILL.md>/../../scripts"; REFS="<...>/../../references"; }

(Windows: python.)

Preflight

python3 "$SCRIPTS/mvsetup.py" --check   # exit 0 = ready; else hand to /motiscope:doctor

Resolve the video from $1, else scan animations/ (as in analyze).

Step 1 — analyze the whole walkthrough

python3 "$SCRIPTS/ingest.py" "<video>" --preset landing

landing gives readable 1280px frames and auto-decomposes onto the motion. The result:

• hold segments = the sections the walkthrough paused on.
• fade-out/cut segments = the transitions between sections.
• per-section entrance timing/easing (the move beats + their cubic-bezier), the stagger timing, and curated frames covering each section.

For a long clip, note it may exceed one pass; you can also re-run focused on a range (--start/--end) to get denser frames for a specific section.

Step 2 — build the site plan (from the frames)