Review animation and motion code against craft standards

This skill reviews animation and motion code against a strict craft, performance, accessibility, and interaction-quality bar, taking a brutal-craft posture: motion that "works" but feels sluggish, lands from the wrong origin, fires too often, or drops frames counts as a regression, not a pass - approval is earned, not assumed. Its substantive standards come from Emil Kowalski's animation philosophy (animations.dev); the review method (non-negotiable standards, escalation triggers, a remedial hierarchy, tiered output, explicit approval criteria) is adapted from aggressive code-quality review. Ten non-negotiable standards gate every animation in a diff: justified motion (must answer "why does this animate?"), frequency-appropriate motion (100+/day or keyboard-triggered actions get no animation), responsive easing (ease-out or a strong custom curve on entering/exiting elements; ease-in on UI is a block), sub-300ms UI duration, correct origin and physicality (popovers scale from their trigger via transform-origin, never from scale(0)), interruptibility (transitions/springs that retarget, not restarting keyframes), GPU-only animated properties (transform/opacity only, not width/height/margin/padding/top/left), prefers-reduced-motion and hover-gating accessibility, asymmetric enter/exit timing for deliberate actions, and cohesion with the product's personality.

Use it when the user asks for an animation, motion, or interaction review, or when a frontend diff changes CSS transitions, keyframes, Framer Motion, WAAPI, hover effects, gestures, toasts, modals, drawers, popovers, or loaders. It reviews motion and animation only - it is not a substitute for a general code review, accessibility audit, or product design critique, does not implement fixes unless separately asked, and declines a request to review general (non-motion) code. Final approval for gestures and highly visual interactions may still need browser, slow-motion, and real-device testing.

A dozen escalation triggers are flagged on sight: transition: all, scale(0) or fade-only entrances with no initial transform, ease-in on UI, animation on a keyboard shortcut or 100+/day action, UI duration over 300ms with no stated reason, transform-origin: center on a trigger-anchored popover, keyframes on rapidly-triggered elements, animating layout properties, a CSS-variable update on a parent driving a child transform (recalc storm), missing reduced-motion handling, ungated hover motion, and symmetric timing on a press-and-hold interaction. Fixes follow a nine-step preference hierarchy from most to least preferred: delete the animation, reduce it, fix the easing, fix origin/physicality, make it interruptible, move it to the GPU, add asymmetric timing, polish (stagger, blur, @starting-style), then accessibility/cohesion. Output is required in two parts: a single markdown findings table (Before/After/Why per row, never a prose before/after list) followed by a verdict section grouped by impact tier (feel-breaking regressions, missed simplifications, performance, interruptibility/timing, origin/physicality/cohesion, accessibility), closing with an explicit Block or Approve decision citing file:line and pulling exact values from STANDARDS.md rather than approximating.

The full rule catalog - easing curves, duration tables, spring configs, gesture handling, clip-path, performance, and a11y specifics - lives in a companion STANDARDS.md, loaded whenever a finding needs a precise value or citation.

Frontend engineers and reviewers who need a rigorous, criteria-based verdict on animation and motion code changes rather than a subjective "looks fine" pass. MIT licensed, by Emil Kowalski.