Motion Scene logo

Motion Scene

Community
Sagargupta16
motion-scene

Designs and builds animated React and Motion (motion/react) scenes for web apps, such as project cover art, service card illustrations, hero art and architecture or mechanism diagrams that show what a thing actually does. Works in three modes, research (read the subject into a storyboard brief), build (implement it with a copy-in TypeScript kit) and verify (browser checks and gates). Use when the user asks for an "animated illustration", "cover scene", "hero animation", "diagram animation in React", "Motion animation for a card", an animated project or service card, a looping mechanism scene, or runs /motion-scene [research|build|verify]. Standalone .svg files go to animated-svg; reviews of existing art go to svg-audit.

Overview

PublisherSagargupta16
Repositorysvg-forge
Skill namemotion-scene
Stars
3
Forks
0
Bundled files
12
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 12 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Sagargupta16 on GitHub. Read the source before you install it.

Installation

Install the Motion Scene AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Sagargupta16/svg-forge/tree/main/skills/motion-scene
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Sagargupta16/svg-forge.git /tmp/svg-forge
mkdir -p .claude/skills
cp -r /tmp/svg-forge/skills/motion-scene .claude/skills/motion-scene
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Motion Scene in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Motion Scene on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Motion Scene is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

motion-scene

A scene is a looping, decorative React component that shows a mechanism (input -> processing -> output) inside a fixed slot of a web app: the media area of a project card, the art column of a service card, a hero panel, a diagram in a docs page. The bar is accuracy. Someone who knows the subject should recognise its mechanism in one glance at card size. A generic "dashboard with bars" is a failure even when it looks good.

Delivery target is scene (a React + Motion component). Timing presets use the shared names micro, short, medium (default), long and ambient.

Which skill

NeedSkill
Animated component rendered inside a React app (card art, hero, diagram, illustration)motion-scene
One self-contained .svg for a README, docs site or <img> (no script, SMIL and CSS only)animated-svg
Review or fix an existing SVG file or scene componentsvg-audit

If the app is not React, or the art must also work as a plain image, use animated-svg. If the user only wants a quick hover or press effect on existing UI, that is a micro interaction, not a scene; apply the spring rules in references/motion.md directly.

Settle these first

Infer from the codebase; ask only what cannot be inferred.

  • Slot: where it renders, its ratio, its smallest and largest rendered width (phone and desktop), and any overlay the host draws on it (bottom fade, badge, title).
  • Tint: the one accent the host passes in (category color, brand accent). Everything else is neutral.
  • Scheme: dark or light canvas.
  • Count: how many instances can be on screen (a grid of 40 cards changes the budget and needs near-viewport mounting).
  • Motion setting: does the app have its own motion control (an in-app Full/Reduced toggle, a settings flag)? Scenes follow it. Never add a reduced-motion gate to an app that has its own motion control. If the app has none, keep motion on and ask before adding one.

Modes

research writes a brief, build implements a brief, verify proves it in the browser. A bare invocation infers the mode from the ask. Always research before building; never build from a README blurb alone.

research

Read, in order: the README, the package manifest (package.json, pyproject.toml, *.tf, *.csproj ...), the entry point, the UI components or CLI command definitions, any screenshots or asset folder. Prefer code over README claims. For a service card with no repo, the truth is what is actually delivered under that heading; depict the deliverable, not the buzzword. For an architecture diagram, read the infrastructure source (IaC, compose files, service definitions), not a slide. For a deployed app, open the live UI (or a screenshot of it) and echo its layout: the scene replaces the screenshot, so a visitor who knows the app should recognise the panel and the one control that matters.

Write the brief before touching code:

markdown
## Brief: <subject>
- What it does: one or two sentences, from the code.
- Mechanism: <input> -> <processing> -> <output>, in the subject's own nouns.
- Vocabulary: 6 to 10 terms that appear verbatim in the source (commands,
  files, tables, models, UI labels). Labels come only from this list.
- Visual identity: what the real UI, CLI or pipeline looks like.
- Current art: none | accurate | generic-but-fitting | partly-wrong | misleading,
  with what it gets right and wrong.
- Strategy: keep | retune-labels | new-variant | new-scene
  (new-scene only when the current art misrepresents the subject).
- Slot: ratio, smallest and largest width, overlays, tint, scheme, count,
  motion setting.
- Preset: medium (3 to 6 s loop) unless the story needs long (8 to 20 s).
- Storyboard, 3 to 6 beats on one clock, as cycle fractions:
  1. 0.04 to 0.22  <what moves, from where to where, which pip or label>
  2. ...
- Budget: <n> animated nodes, <m> labels (see composition budgets).
- Evidence: file and line for every claim above.

Each beat must be buildable with absolutely positioned SVG and HTML at the smallest slot width. If a beat only reads at hero size, cut it or move it to a long storyboard.

build

  1. Copy the kit into the app (for example src/scenes/kit/): tokens.ts, timing.ts, routes.ts, primitives.tsx, useNearViewport.ts, useFreezeAnimations.ts. It needs React 19 and motion. If the slot is not 16:10, change STAGE_W and STAGE_H in tokens.ts together. Read references/primitives.md for what each part does.
  2. Start from the storyboard, not from another scene's shapes. Place parts in stage units (320 x 200 for 16:10). SVG geometry uses the stage viewBox; HTML packets, pips and labels sit at stage percentages with fixed px sizes. See references/composition.md.
  3. One clock. const beat = clock(PRESET.medium) and every loop goes through beat(...). Times are cycle fractions. Derive event times (passAt) instead of hand-tuning them. Choose the loop shape: build up to one END/GONE exit, or open on the finished state and clear in the first 4 to 8 percent. Write the beat table as a comment at the top of the file. See references/motion.md.
  4. Compose from the primitives: Shell (root, palette, texture, tint wash, stage), Panel and Wire in the SVG layer, Trace for animated strokes, Packet with ride and comet for data in flight, Pip for status, Label for the 3 or 4 words, FadeLayer to fade a group as one node. Custom motion elements inside a Shell take useLoop(loop) so they freeze with the rest.
  5. Variants, not forks. Subjects that share a mechanism get one scene family with a variant prop, dispatched through a Record with a real fallback. A variant must change a moving element, not only label text.
  6. Mount through a host slot: a fixed-ratio box, the scene lazy-loaded, mounted only near the viewport (useNearViewport), a placeholder painted with palette.base otherwise, frozen wired to the app's motion setting, useFreezeAnimations as the backstop, and the scene keyed on the setting. SceneSlot in ExampleScene.tsx is the pattern.
  7. Keep it lint-clean by construction (readonly props, no nested ternaries, no nested template literals, hoisted repeated literals, small sub-components). One default export per scene file; split past about 450 lines. See references/gotchas.md.
  8. Run the app's own gates (lint, type-check, format, tests, build), then verify.

ExampleScene.tsx is a complete 6 s read-through cache loop built only from the kit: 8 animated nodes, 3 labels, two packets on one route, derived miss, fill and hit beats.

Scene contract:

  • Props Readonly<{ tint: string; frozen?: boolean }>, plus variant?: string only in a family that reads it.
  • Root is Shell: absolute, inset 0, clipped, aria-hidden="true". The host card's title and text carry the meaning.
  • Default export, lazy-loaded by the host so scenes stay out of the entry chunk.

verify

Run the app, open the page in a focused tab after a hard reload, and work through references/verify.md. The short version:

  • Screenshot at the smallest and largest real slot widths; the mechanism reads at both.
  • Animated node count within budget; running animations in view under about 70.
  • Easing scan returns 0; every loop shares one duration.
  • Two full cycles watched: no seam jump. Transforms sampled twice 700 ms apart differ.
  • No horizontal overflow at 375 px; paint flashing shows the card body is not repainting.
  • Root is aria-hidden; no decoration sits over text or controls.
  • With the app's reduced setting on, an 800 ms computed-style diff shows 0 changes and the artwork is still visible.
  • The app's gates pass. One scene family per commit.

Report each check with its measured value, not "looks fine".

Hard rules

  1. Mechanism, not category. Input -> processing -> output in the subject's own nouns. Changing label text does not make a scene accurate. composition
  2. Labels: at most 4 (2 on icon-size cards), uppercase mono, 7 to 9 px, 12 characters or fewer, no dashes, real vocabulary only, rendered as HTML so they keep their px size. No prose and no numbers that read as claims. composition
  3. Palette is the tint plus neutrals. Alpha ladders of one tint and one neutral; ok green only for live or success, warn amber only for warnings. No new hue, gradient text, glow halo or blur filter. composition
  4. One stage at the slot ratio: uniform viewBox, HTML at stage percentages with fixed px sizes, letterboxed by a container-query stage. Never stretch with preserveAspectRatio="none" unless the viewBox matches the slot exactly. composition
  5. Budgets: 12 or fewer animated nodes per cover, 8 or fewer per icon-size card, 3 to 6 beats, under about 70 running animations in view; mount loops only near the viewport. composition
  6. Transform and opacity only (plus SVG pathLength and pathOffset for stroke draws). Never animate width, height, top, left, filter or box-shadow. motion
  7. One clock per scene. One duration for every loop, events as cycle fractions; ambient backdrops may run slower on their own speed. motion
  8. Per-segment easing whenever times is set: ease: perSegment(times). A single ease string drifts WAAPI tracks off the storyboard. motion
  9. Seamless loops: last keyframe equals the first, or the reset happens while the element is invisible; one END/GONE exit per scene, or open on the finished state with a clear window. motion
  10. Never put pathLength or pathOffset on a path with vector-effect="non-scaling-stroke". Static hairlines (Wire) and animated draws (Trace) are separate elements. gotchas
  11. Decoration only: root aria-hidden, nothing focusable inside, never over text or controls. verify
  12. Follow the app's motion setting. Freeze through Shell frozen and useFreezeAnimations; never add an OS-level gate to an app that has its own control. Motion stays visible by default. gotchas
  13. Never start from scale(0); start at 0.9 to 0.97 (0.25 for icons). Exits run faster than entrances. motion

Kit files

FileWhat it holds
tokens.tsStage size, pctX/pctY, createPalette (tint, alpha ladders, base, solid), textures, dotAt
timing.tsPRESET, clock, perSegment, loopProps, shown, windowTimes/windowKeys, ambient
routes.tsroute, sCurve, line, ride, comet, lit, passAt
primitives.tsxShell, Stage, Panel, Wire, Trace, Packet, Pip, Label, FadeLayer, useLoop, useScene
useNearViewport.tsMount a scene only within 300 px of the viewport
useFreezeAnimations.tsPause every WAAPI animation in a subtree at time 0
ExampleScene.tsxComplete 6 s mechanism loop plus the SceneSlot host pattern

The kit type-checks under strict, noUncheckedIndexedAccess and exactOptionalPropertyTypes with React 19 and Motion 14.

References

  • references/composition.md: mechanism over category, labels, palette, depth layers, stage units and responsive layout, budgets, variants.
  • references/motion.md: presets mapped to scene use, one clock, cycle fractions, per-segment easing and the WAAPI drift, seamless loops, stagger, springs vs tweens, enter and exit.
  • references/gotchas.md: every known failure as symptom, cause, fix.
  • references/primitives.md: each kit part, when to use it, short snippets.
  • references/verify.md: screenshot sizes, budgets, scan snippets, seam check, paint flashing, accessibility, reduced mode, gates.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Motion Scene AI skill do?

Designs and builds animated React and Motion (motion/react) scenes for web apps, such as project cover art, service card illustrations, hero art and architecture or mechanism diagrams that show what a thing actually does. Works in three modes, research (read the subject into a storyboard brief), build (implement it with a copy-in TypeScript kit) and verify (browser checks and gates). Use when the user asks for an "animated illustration", "cover scene", "hero animation", "diagram animation in React", "Motion animation for a card", an animated project or service card, a looping mechanism scen...

Why use Motion Scene on TypingMind?

Because you install it once and use it with any model. Motion Scene is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Motion Scene in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Sagargupta16/svg-forge/tree/main/skills/motion-scene. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Motion Scene?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Motion Scene?

As many as you like. As long as a model supports skills, you can use Motion Scene with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Motion Scene AI skill free?

Yes. It is published on GitHub by Sagargupta16 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇