Advanced Compose visuals - Material 3 Expressive motion physics, AGSL shaders (Android 13+), Canvas/DrawScope generative, graphicsLayer effects.
npx skills add https://github.com/AThevon/genjutsu --skill compose-graphics
> Advanced Compose visuals: M3 Expressive motion physics, AGSL shaders (Android 13+), Canvas / DrawScope, graphicsLayer effects.
> Loaded for advanced thesis (shader, expressive, M3 Expressive, AGSL, Canvas, holographic).
> Foundation: ../compose-motion/SKILL.md covers basics. Concise rules here. Deep-dive in references/.
| Need | API |
|---|---|
| Spring physics with bounce / overshoot | MotionScheme.expressive() (M3 Expressive) |
| Pixel-level shader | RuntimeShader + Modifier.graphicsLayer { renderEffect = ... } (Android 13+) |
| Generative drawing (paths, particles, fractals) | Canvas { drawScope -> ... } |
| GPU effects (blur, shadows, color filters) | Modifier.graphicsLayer { renderEffect = ... } or Modifier.blur(...) |
| Adaptive system materials (Material You glassmorphism) | Modifier.background(MaterialTheme.colorScheme.surfaceContainerHighest) |
| Liquid glass on Android | AGSL shader recipe (no native API like iOS yet) |
The 2025 Material 3 evolution introduces spring-based motion physics replacing fixed-duration tweens. New shape morphing API via androidx.graphics.shapes. New MotionScheme selectable on the theme. Aimed at hero moments, key interactions, brand-defining UI.
| Scheme | Personality | Use For |
|---|---|---|
| MotionScheme.standard() | Calmer, less overshoot | Default for chrome, lists, navigation |
| MotionScheme.expressive() | More overshoot, longer settle | Hero reveals, FABs, primary CTAs |
Apply on the theme:
MaterialTheme(motionScheme = MotionScheme.expressive()) {
// children read tokens via MaterialTheme.motionScheme.*
}
Tokens exposed:
| Token | Domain | Speed |
|---|---|---|
| fastSpatialSpec() | Position / size | < 200ms |
| defaultSpatialSpec() | Position / size | ~ 350ms |
| slowSpatialSpec() | Position / size | ~ 600ms |
| fastEffectsSpec() | Opacity / color | < 150ms |
| defaultEffectsSpec() | Opacity / color | ~ 250ms |
| slowEffectsSpec() | Opacity / color | ~ 400ms |
> Spatial vs Effects: spatial = anything physical (height, offset, scale). Effects = visual properties without inertia (alpha, color, elevation). Springs feel natural for spatial; tweens feel right for effects. The tokens encode this for you.
@Composable
fun ExpressiveHero() {
var expanded by remember { mutableStateOf(false) }
MaterialTheme(motionScheme = MotionScheme.expressive()) {
val transition = updateTransition(targetState = expanded, label = "expand")
val height by transition.animateDp(
transitionSpec = { MaterialTheme.motionScheme.slowSpatialSpec() },
label = "height"
) { if (it) 400.dp else 100.dp }
val alpha by transition.animateFloat(
transitionSpec = { MaterialTheme.motionScheme.defaultEffectsSpec() },
label = "alpha"
) { if (it) 1f else 0f }
Card(
modifier = Modifier
.fillMaxWidth()
.height(height)
.clickable { expanded = !expanded }
) {
Box(modifier = Modifier.alpha(alpha)) {
Text("Detail content", modifier = Modifier.padding(24.dp))
}
}
}
}
| Mood | Spec |
|---|---|
| Hero reveal | spring(stiffness = Spring.StiffnessLow, dampingRatio = Spring.DampingRatioMediumBouncy) |
| Snappy expressive | spring(stiffness = Spring.StiffnessMediumLow, dampingRatio = 0.7f) |
| Calm spatial | MaterialTheme.motionScheme.defaultSpatialSpec() (use the token) |
| Critical (no overshoot) | spring(stiffness = Spring.StiffnessHigh, dampingRatio = 1f) |
androidx.graphics.shapes ships predefined morphable shapes (MaterialShapes.Circle, Pentagon, Cookie4Sided, Sunny, Heart, etc.) and a Morph(start, end) interpolator.
val morph = remember { Morph(MaterialShapes.Circle, MaterialShapes.Cookie4Sided) }
val progress by animateFloatAsState(
targetValue = if (active) 1f else 0f,
animationSpec = MaterialTheme.motionScheme.slowSpatialSpec(),
label = "morph"
)
Box(
modifier = Modifier
.size(96.dp)
.clip(GenericShape { size, _ ->
addPath(
morph.toPath(progress).asAndroidPath().asComposePath()
)
})
.background(MaterialTheme.colorScheme.primary)
)
AGSL is Android's shader language. Similar to GLSL with simplifications (half4 instead of vec4, restricted feature set, sandbox-safe). RuntimeShader compiles your AGSL source. Bind to a Compose modifier via Modifier.graphicsLayer { renderEffect = ... }.
@Composable
fun ShaderEffect(content: @Composable () -> Unit) {
if (Build.VERSION.SDK_INT < Build.VERSION_CODES.TIRAMISU) {
// Android 12 and below: skip the shader, render content as-is.
content()
return
}
val shader = remember { RuntimeShader(AGSL_SOURCE) }
val time by produceState(0f) {
while (true) {
withFrameMillis { ms ->
value = ms / 1000f
}
}
}
Box(
modifier = Modifier
.onSizeChanged {
shader.setFloatUniform("resolution", it.width.toFloat(), it.height.toFloat())
}
.graphicsLayer {
shader.setFloatUniform("time", time)
renderEffect = RenderEffect
.createRuntimeShaderEffect(shader, "image")
.asComposeRenderEffect()
}
) { content() }
}
> Always gate on Build.VERSION.SDK_INT >= 33. Pre-Android 13 fallback: render the content unmodified, or use Modifier.blur() / a static gradient overlay.
AGSL source:
// ripple.agsl
uniform float2 resolution;
uniform float2 origin;
uniform float time;
uniform shader image;
half4 main(float2 fragCoord) {
float2 toOrigin = fragCoord - origin;
float dist = length(toOrigin);
float wave = sin(dist * 0.05 - time * 8.0) * 0.05;
float falloff = 1.0 / max(dist * 0.01, 1.0);
float2 dir = toOrigin / max(dist, 0.0001);
float2 displaced = fragCoord + dir * wave * falloff * 50.0;
return image.eval(displaced);
}
Compose binding (state-driven time, animated via LaunchedEffect):
@Composable
fun RippleSurface(content: @Composable BoxScope.() -> Unit) {
if (Build.VERSION.SDK_INT < 33) { Box { content() }; return }
val shader = remember { RuntimeShader(RIPPLE_AGSL) }
var origin by remember { mutableStateOf(Offset.Zero) }
var startMs by remember { mutableStateOf<Long?>(null) }
val time by produceState(0f, startMs) {
if (startMs == null) { value = 0f; return@produceState }
val begin = startMs!!
while (true) {
withFrameMillis { now ->
value = (now - begin) / 1000f
}
}
}
Box(
modifier = Modifier
.onSizeChanged {
shader.setFloatUniform("resolution", it.width.toFloat(), it.height.toFloat())
}
.pointerInput(Unit) {
detectTapGestures { tap ->
origin = tap
startMs = System.currentTimeMillis()
}
}
.graphicsLayer {
shader.setFloatUniform("origin", origin.x, origin.y)
shader.setFloatUniform("time", time)
renderEffect = RenderEffect
.createRuntimeShaderEffect(shader, "image")
.asComposeRenderEffect()
},
content = content
)
}
uniform float time;
uniform shader image;
half4 main(float2 fragCoord) {
half4 color = image.eval(fragCoord);
float n = fragCoord.x * 0.01 + fragCoord.y * 0.005 + time * 0.3;
half3 rainbow = half3(
sin(n * 2.0) * 0.5 + 0.5,
sin(n * 2.0 + 2.094) * 0.5 + 0.5,
sin(n * 2.0 + 4.188) * 0.5 + 0.5
);
half luma = dot(color.rgb, half3(0.299, 0.587, 0.114));
return half4(mix(color.rgb, rainbow * luma * 2.0, 0.5), color.a);
}
Multiplies the underlying view's luminance by a phase-shifted RGB sine wave - shimmer over text or icons. time animates the wave drift.
Android lacks a native iOS-26-style glassEffect. Compose this with blur + chromatic aberration + soft edge tint:
uniform float2 resolution;
uniform shader image;
half4 main(float2 fragCoord) {
float2 uv = fragCoord / resolution;
// Chromatic aberration: sample R, G, B at slightly different offsets
float2 ca = (uv - 0.5) * 0.004;
half r = image.eval(fragCoord + ca * resolution).r;
half g = image.eval(fragCoord).g;
half b = image.eval(fragCoord - ca * resolution).b;
half4 base = half4(r, g, b, 1.0);
// Soft edge tint: brighten near borders for that frosted look
float edge = smoothstep(0.45, 0.5, max(abs(uv.x - 0.5), abs(uv.y - 0.5)));
half3 tinted = base.rgb + half3(edge * 0.08);
return half4(tinted, 0.85);
}
Pair with Modifier.blur(20.dp, BlurredEdgeTreatment.Unbounded) upstream for the actual blur (cheaper than computing it in AGSL).
RuntimeShader is a render pass. Compounds at 60fps - benchmark via Macrobenchmark.image.eval(coord) samples the underlying view as a texture. Heavy if the view is complex (long lists, nested layouts). Cache static parts in a parent that doesn't recompose.Canvas(modifier = Modifier.size(200.dp)) {
// 'this' is DrawScope - size, drawCircle, drawPath, drawRect...
drawCircle(Color.Blue, radius = size.minDimension / 2)
drawPath(myPath, color = Color.White, style = Stroke(width = 4.dp.toPx()))
}
| DrawScope method | Use |
|---|---|
| drawCircle | Filled or stroked circles |
| drawRect | Rectangles, optionally with gradient brush |
| drawPath | Arbitrary path with Stroke or Fill |
| drawArc | Slice of an oval - useful for progress arcs |
| drawLine | Straight segment between two Offset |
| drawText | Text via TextMeasurer (preferred) or drawIntoCanvas { it.nativeCanvas.drawText(...) } |
| drawIntoCanvas { } | Escape hatch - access raw android.graphics.Canvas |
@Composable
fun AnimatedWave() {
val infinite = rememberInfiniteTransition(label = "wave")
val phase by infinite.animateFloat(
initialValue = 0f,
targetValue = 2 * PI.toFloat(),
animationSpec = infiniteRepeatable(
animation = tween(3000, easing = LinearEasing)
),
label = "phase"
)
Canvas(modifier = Modifier.fillMaxWidth().height(80.dp)) {
val path = Path()
var x = 0
while (x <= size.width.toInt()) {
val y = size.height / 2 + sin(x * 0.05f + phase) * 20f
if (x == 0) path.moveTo(x.toFloat(), y) else path.lineTo(x.toFloat(), y)
x += 2
}
drawPath(path, color = Color.Blue, style = Stroke(width = 2.dp.toPx()))
}
}
data class Particle(var x: Float, var y: Float, var vx: Float, var vy: Float, var life: Float)
@Composable
fun ParticleField() {
val particles = remember {
mutableStateListOf<Particle>().apply {
repeat(50) {
add(Particle(
x = Random.nextFloat() * 1000f,
y = Random.nextFloat() * 1000f,
vx = (Random.nextFloat() - 0.5f) * 4f,
vy = (Random.nextFloat() - 0.5f) * 4f,
life = 1f
))
}
}
}
LaunchedEffect(Unit) {
while (true) {
withFrameMillis { /* tick */ }
particles.forEachIndexed { i, p ->
p.x += p.vx
p.y += p.vy
p.life -= 0.01f
if (p.life <= 0f) {
p.x = Random.nextFloat() * 1000f
p.y = Random.nextFloat() * 1000f
p.life = 1f
}
}
}
}
Canvas(modifier = Modifier.fillMaxSize()) {
particles.forEach { p ->
drawCircle(
color = Color.White.copy(alpha = p.life),
radius = 2.dp.toPx(),
center = Offset(p.x, p.y)
)
}
}
}
Quick mention: noise-driven particle motion. See references/canvas-generative.md for the full deep-dive (Perlin noise, particle pools, advection loop).
derivedStateOf to gate redraws.LaunchedEffect(Unit) { while(true) { withFrameMillis { ... } } } is fine for 60fps. Avoid delay(16) - it drifts.remember, only animate transforms via translate { drawPath(...) }.// BAD - every interaction overshoots, UI feels like a bouncy castle
MaterialTheme(motionScheme = MotionScheme.expressive()) {
AppRoot()
}
// GOOD - Standard for chrome, scope Expressive to hero moments
MaterialTheme(motionScheme = MotionScheme.standard()) {
NavigationScaffold {
// Hero detail screen overrides locally
MaterialTheme(motionScheme = MotionScheme.expressive()) {
HeroDetail()
}
}
}
// BAD - crashes on Android 12 and below
val shader = remember { RuntimeShader(AGSL_SOURCE) }
// GOOD - gate on SDK level, provide fallback
if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.TIRAMISU) {
ShaderEffect { content() }
} else {
Box(modifier = Modifier.background(fallbackGradient)) { content() }
}
// BAD - any state change in the parent triggers a recompose of the Canvas
Canvas(modifier = Modifier.fillMaxSize()) {
drawCircle(color = if (viewModel.isActive) Color.Red else Color.Blue, radius = 50f)
}
// GOOD - hoist the reads, derive a stable value
val color by remember { derivedStateOf { if (viewModel.isActive) Color.Red else Color.Blue } }
Canvas(modifier = Modifier.fillMaxSize()) {
drawCircle(color = color, radius = 50f)
}
// BAD - 4 render passes per frame, GPU melts on mid-range devices
.graphicsLayer { renderEffect = blurEffect }
.graphicsLayer { renderEffect = chromaticEffect }
.graphicsLayer { renderEffect = noiseEffect }
.graphicsLayer { renderEffect = vignetteEffect }
// GOOD - one shader doing all the math in a single pass
.graphicsLayer { renderEffect = combinedGlassEffect }
// BAD - new Path every frame, GC stutter
Canvas(modifier = Modifier.fillMaxSize()) {
val path = Path()
points.forEach { path.lineTo(it.x, it.y) }
drawPath(path, color = Color.Black)
}
// GOOD - reuse a remembered Path, rewind each frame
val path = remember { Path() }
Canvas(modifier = Modifier.fillMaxSize()) {
path.rewind()
points.forEach { path.lineTo(it.x, it.y) }
drawPath(path, color = Color.Black)
}
| Need | Load |
|---|---|
| AGSL recipes (7 working shaders with binding code) | references/agsl-recipes.md |
| M3 Expressive choreography deep-dive | references/m3-expressive-deep.md |
| Generative drawing patterns (flow fields, L-systems, particles) | references/canvas-generative.md |
| Base animations (animateAsState, AnimatedVisibility, Transition) | ../compose-motion/SKILL.md |
| CMP patterns (shared UI iOS / Android / Desktop) | ../compose-multiplatform/SKILL.md |
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Improves the quality of images, especially screenshots, by enhancing resolution, sharpness, and clarity. Perfect for preparing images for presentations, documentation, or social media posts.
Downloads videos from YouTube and other platforms for offline viewing, editing, or archival. Handles various formats and quality options.
Lightweight WSI tile extraction and preprocessing. Use for basic slide processing tissue detection, tile extraction, stain normalization for H&E images. Best for simple pipelines, dataset preparation, quick tile-based analysis. For advanced spatial proteomics, multiplexed imaging, or deep learning pipelines use pathml.
Microscopy data management platform. Access images via Python, retrieve datasets, analyze pixels, manage ROIs/annotations, batch processing, for high-content screening and microscopy workflows.
Python library for working with DICOM (Digital Imaging and Communications in Medicine) files. Use this skill when reading, writing, or modifying medical imaging data in DICOM format, extracting pixel data from medical images (CT, MRI, X-ray, ultrasound), anonymizing DICOM files, working with DICOM metadata and tags, converting DICOM images to other formats, handling compressed DICOM data, or processing medical imaging datasets. Applies to tasks involving medical image analysis, PACS systems, radiology workflows, and healthcare imaging applications.
This skill should be used when working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks. Use for text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning models on custom datasets.
Take athevon/compose-graphics from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.