Makepad 2.0 Shaders logo

Makepad 2.0 Shaders

Community
ZhangHanDong
makepad-2.0-shaders

CRITICAL: Use for Makepad 2.0 shader system. Triggers on: makepad shader, Sdf2d, pixel shader, draw_bg, draw_text, draw_quad, makepad gpu, shader function, pixel fn, vertex fn, instance, uniform, shader variable, sdf, premultiply, Pal.premul, GaussShadow, makepad graphics, custom draw, DrawQuad, DrawVector, 着色器, 像素, 渲染, 自定义绘制, 距离场

Overview

PublisherZhangHanDong
Repositorymakepad-skills
Skill namemakepad-2.0-shaders
Stars
748
Forks
88
Bundled files
2
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.

  • 2 bundled files

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

  • Open source

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

Installation

Install the Makepad 2.0 Shaders 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/ZhangHanDong/makepad-skills/tree/main/skills/makepad-2.0-shaders
  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/ZhangHanDong/makepad-skills.git /tmp/makepad-skills
mkdir -p .claude/skills
cp -r /tmp/makepad-skills/skills/makepad-2.0-shaders .claude/skills/makepad-2.0-shaders
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Makepad 2.0 Shaders 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 Makepad 2.0 Shaders 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 Makepad 2.0 Shaders 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.

Makepad 2.0 Shader Skill

Version: makepad-widgets (dev branch) | Last Updated: 2026-03-03

Overview

Makepad uses a custom GPU shader system integrated into the widget property tree. Shaders are defined inline using pixel: fn() { ... } and vertex: fn() { ... } blocks within draw_bg, draw_text, or custom draw objects.

Documentation

Refer to the local files for detailed documentation:

  • ./references/shader-reference.md - Shader syntax, variables, built-ins, custom functions
  • ./references/sdf2d-reference.md - SDF2D primitives, combinators, drawing operations

Shader Basics

Pixel Shader Structure

draw_bg +: {
    // Declare variables
    instance hover: 0.0          // Animatable per-instance
    uniform accent: #4488ff      // Shared across all instances

    pixel: fn() {
        let sdf = Sdf2d.viewport(self.pos * self.rect_size)
        // ... SDF operations ...
        return sdf.result
    }
}

Variable Types

TypeDeclarationAnimatableScope
instanceinstance hover: 0.0Yes (via Animator)Per-widget instance
uniformuniform color: #fffNoShared across instances
texture_2dtexture_2d tex: noneNoTexture sampler
varyingvarying uv: vec2NoVertex → fragment

Built-in Variables

VariableTypeDescription
self.posvec2Normalized position (0.0 to 1.0)
self.rect_sizevec2Widget size in pixels
self.dpi_factorfloatScreen DPI factor
self.draw_pass.timefloatTime in seconds

CRITICAL: Premultiply Alpha

Every pixel shader MUST return premultiplied alpha color!

// WRONG - non-premultiplied
pixel: fn() {
    return vec4(1.0, 0.0, 0.0, 0.5)
}

// CORRECT - use Pal.premul()
pixel: fn() {
    return Pal.premul(vec4(1.0, 0.0, 0.0, 0.5))
}

// ALSO CORRECT - sdf.result is already premultiplied
pixel: fn() {
    let sdf = Sdf2d.viewport(self.pos * self.rect_size)
    sdf.circle(cx, cy, r)
    sdf.fill(#f00)
    return sdf.result
}

SDF2D Quick Reference

Setup

let sdf = Sdf2d.viewport(self.pos * self.rect_size)

Primitives

sdf.circle(cx, cy, radius)
sdf.rect(x, y, w, h)
sdf.box(x, y, w, h, border_radius)
sdf.hexagon(cx, cy, radius)
sdf.arc(cx, cy, radius, start_angle, end_angle, thickness)
sdf.move_to(x, y)
sdf.line_to(x, y)
sdf.close_path()

Drawing

sdf.fill(color)         // Filled shape
sdf.stroke(color, width) // Outlined shape
sdf.glow(color, amount)  // Glow effect
sdf.clear(color)         // Clear with color

Combinators

sdf.union()       // Add shapes together
sdf.intersect()   // Keep overlap only
sdf.subtract()    // Remove second from first
sdf.gloop(radius) // Smooth union
sdf.blend(amount) // Linear blend

Transforms

sdf.translate(x, y)
sdf.rotate(angle, cx, cy)
sdf.scale(factor, cx, cy)

Color Operations

// Mix two colors
mix(#f00, #00f, 0.5)           // 50% blend

// Premultiply alpha
Pal.premul(vec4(r, g, b, a))

// HSV conversions
Pal.hsv2rgb(vec4(h, s, v, 1.0))
Pal.rgb2hsv(color)

// Random
Math.random_2d(vec2(x, y))

Common Shader Patterns

Gradient Background

draw_bg +: {
    pixel: fn() {
        let grad = mix(#1a1a2e, #16213e, self.pos.y)
        return Pal.premul(vec4(grad.xyz, 1.0))
    }
}

Hover Color Change

draw_bg +: {
    instance hover: 0.0
    color: #333
    pixel: fn() {
        return Pal.premul(mix(self.color, self.color * 1.3, self.hover))
    }
}

Box Shadow

draw_bg +: {
    pixel: fn() {
        let sdf = Sdf2d.viewport(self.pos * self.rect_size)
        // Shadow
        sdf.box(2.0, 2.0, self.rect_size.x - 4.0, self.rect_size.y - 4.0, 8.0)
        sdf.fill(GaussShadow.box_shadow(sdf, 4.0, #0005))
        // Card
        sdf.box(0.0, 0.0, self.rect_size.x - 2.0, self.rect_size.y - 2.0, 8.0)
        sdf.fill(#2a2a3d)
        return sdf.result
    }
}

Rounded Button with States

draw_bg +: {
    instance hover: 0.0
    instance down: 0.0
    uniform color_bg: #4488ff
    uniform color_hover: #5599ff
    uniform color_down: #3377ee

    pixel: fn() {
        let sdf = Sdf2d.viewport(self.pos * self.rect_size)
        sdf.box(0.0, 0.0, self.rect_size.x, self.rect_size.y, 6.0)
        let color = mix(self.color_bg, self.color_hover, self.hover)
        let color = mix(color, self.color_down, self.down)
        sdf.fill(color)
        return sdf.result
    }
}

Custom Shader Functions

draw_bg +: {
    fn wave(pos: vec2, time: float) -> float {
        return sin(pos.x * 10.0 + time * 3.0) * 0.1
    }

    pixel: fn() {
        let w = self.wave(self.pos, self.draw_pass.time)
        let color = mix(#1a1a2e, #4488ff, self.pos.y + w)
        return Pal.premul(vec4(color.xyz, 1.0))
    }
}

Splash Shader Capability Boundary

Splash CAN:

  • Override pixel: fn(), vertex: fn(), get_color: fn() on existing draw types via +:
  • Define helper shader functions within +: blocks
  • Set instance/uniform variables on existing draw types
  • Use all SDF2D, color, math built-ins in shader functions

Splash CANNOT:

  • Create new DrawQuad/DrawText/DrawSvg types (must define in Rust)
  • Add new instance fields to existing shaders (GPU layout is compile-time)

Rule: Rust defines the draw type struct + registers it; Splash overrides how it draws.

See ./references/shader-reference.md "Splash Shader Capabilities & Boundaries" for the full pattern.


Custom Fullscreen Shader Pattern (learned 2026-03-26)

For standalone shader-driven widgets (e.g. particle fields, visualizers), follow the examples/shader pattern:

1. Custom Draw Type (Rust)

rust
#[derive(Script, ScriptHook)]
#[repr(C)]  // CRITICAL: must be repr(C) for GPU layout
pub struct DrawMyShader {
    #[deref] draw_super: DrawQuad,  // inherits from DrawQuad
    #[live] my_param: f32,           // maps to shader variable
}

2. Register + Define Shader (script_mod!)

set_type_default() do #(DrawMyShader::script_shader(vm)){
    ..mod.draw.DrawQuad        // inherit DrawQuad defaults
    my_param: 0.5              // default value

    // Custom functions: property-style syntax, NOT fn name(self, ...)
    my_helper: fn(a: float, b: float) -> vec2 {
        return vec2(a * 2.0, b * 0.5)
    }

    pixel: fn() {
        let result = self.my_helper(self.pos.x, self.pos.y)
        return Pal.premul(vec4(result.x, result.y, 0.0, 1.0))
    }
}

3. Widget with Turtle Layout

rust
fn draw_walk(&mut self, cx: &mut Cx2d, _: &mut Scope, walk: Walk) -> DrawStep {
    cx.begin_turtle(walk, self.layout);
    let rect = cx.turtle().rect();
    self.draw_bg.draw_abs(cx, rect);      // single fullscreen quad
    cx.end_turtle_with_area(&mut self.area);
    DrawStep::done()
}

4. Updating Shader Variables from Rust

rust
// Direct field access (when draw type has #[live] fields):
self.draw_bg.my_param = 0.75;
self.area.redraw(cx);

// Via NextFrame for animation:
if let Event::NextFrame(ne) = event {
    if ne.set.contains(&self.next_frame) {
        self.draw_bg.my_param += 0.01;
        self.area.redraw(cx);
        self.next_frame = cx.new_next_frame();
    }
}

Instanced Particle Rendering (learned 2026-03-26)

For drawing thousands of independent particles (dots, stars, etc.):

Draw Shader

rust
#[derive(Script, ScriptHook)]
#[repr(C)]
pub struct DrawDot {
    #[deref] draw_super: DrawQuad,
    #[live] dot_color: Vec3,  // per-instance color
}
// Shader: each instance is a small circle
pixel: fn() {
    let d = length(self.pos - vec2(0.5, 0.5))
    let alpha = 1.0 - smoothstep(0.35, 0.5, d)
    return Pal.premul(vec4(self.dot_color * alpha, alpha))
}

Rendering Loop

rust
self.draw_dot.begin_many_instances(cx);  // start batch

for i in 0..particles.len() {
    let (x, y) = particles[i];
    self.draw_dot.dot_color = vec3(r, g, b);  // set per-instance data
    self.draw_dot.draw_abs(cx, Rect {
        pos: dvec2(x - radius, y - radius),
        size: dvec2(radius * 2.0, radius * 2.0),
    });
}

self.draw_dot.end_many_instances(cx);  // submit batch as one draw call

Physics Pattern (spring-back displacement)

rust
// Per particle: store persistent displacement
displacements: Vec<(f64, f64)>,

// Each frame:
for i in 0..dots.len() {
    let (mut dx, mut dy) = displacements[i];

    // 1. Decay (spring back, 0.94 = ~2-3 sec return)
    dx *= 0.94;
    dy *= 0.94;

    // 2. Apply forces (cursor push, ripples, etc.)
    let dist = distance(dot_pos, mouse_pos);
    let t = (1.0 - dist / radius).max(0.0);
    let push = t * t * t * strength;  // cubic falloff
    dx += direction.x * push;
    dy += direction.y * push;

    displacements[i] = (dx, dy);
    // Draw at original_pos + displacement
}

Performance Notes

  • 10,000 particles at 60fps: OK on macOS Metal (one draw call via instancing)
  • CPU physics loop: 10K × 17 distance checks = ~170K ops/frame, negligible
  • Key: begin_many_instances / end_many_instances batches into single GPU draw call

Shader Syntax Pitfalls (learned 2026-03-26)

PitfallErrorFix
let x = 1.0; x = 2.0cannot assign to let bindingUse different names: let x2 = ...
fn push(self, ...) -> vec2method not found on selfUse property syntax: push: fn(...) -> vec2 { }
return vec4(r, g, b, a) without premulIncorrect alpha blendingreturn Pal.premul(vec4(r, g, b, a))
Custom shader in Splash evalSilent blank renderMust use compiled script_mod! path
Missing #[repr(C)] on draw structGPU layout mismatchAlways add #[repr(C)]
fn calc(self, x: float) syntaxcannot push to frozen vecUse calc: fn(x: float) -> float { }

Best Practices

  1. Always premultiply - Use Pal.premul() or return sdf.result
  2. Use instance for animation - Only instance variables work with Animator
  3. Use uniform for shared values - Colors, sizes shared across instances
  4. Use +: merge operator - Extend default shaders: draw_bg +: { ... }
  5. Keep shaders simple - Complex shaders impact rendering performance
  6. Use SDF for shapes - Much cleaner than manual math
  7. Test with new_batch: true - Required when mixing shaders with text
  8. Property-style functions - name: fn(args) -> type { }, call via self.name(args)
  9. Immutable let - Shader let cannot be reassigned; use unique names per step
  10. #[repr(C)] on draw structs - Required for GPU memory layout alignment

SDF Capsule (Pill) Shape Pattern (learned 2026-03-31)

sdf.box() with large border_radius breaks when radius approaches half the dimension — the formula size.xy - vec2(2*r, 2*r) goes negative, producing diamond/spiky shapes. Use this standard capsule SDF instead:

draw_bg +: {
    pixel: fn() {
        let w = self.rect_size.x
        let h = self.rect_size.y
        let r = h * 0.5
        let px = self.pos.x * w
        let py = self.pos.y * h
        // Standard capsule: clamp x to center segment, then circle distance
        let cx = clamp(px, r, max(r, w - r))
        let cy = h * 0.5
        let d = length(vec2(px - cx, py - cy)) - r
        let alpha = 1.0 - smoothstep(-1.0, 1.0, d)
        return Pal.premul(vec4(0.1, 0.1, 0.18, alpha * 0.82))
    }
}

Key points:

  • clamp(px, r, w-r) constrains x to the center line segment between the two end circles
  • max(r, w-r) prevents clamp range inversion when widget is very narrow
  • Smoothstep (-1.0, 1.0, d) provides 2px anti-aliasing
  • Works correctly at any width — dynamically adapts as width: Fit content changes
  • No three-part union (circles + rect) needed — single formula, no seam artifacts

Pulsing Dot in Background Shader (learned 2026-03-31)

Embed animation directly in the background shader to avoid z-order issues with child widgets (LoadingSpinner/other widgets can cause bleed-through at capsule edges):

draw_bg +: {
    pixel: fn() {
        let w = self.rect_size.x
        let h = self.rect_size.y
        let r = h * 0.5
        let px = self.pos.x * w
        let py = self.pos.y * h

        // Capsule background
        let cx_bg = clamp(px, r, max(r, w - r))
        let cy = h * 0.5
        let d_bg = length(vec2(px - cx_bg, py - cy)) - r
        let bg_alpha = 1.0 - smoothstep(-1.0, 1.0, d_bg)
        let bg = vec4(0.1, 0.1, 0.18, bg_alpha * 0.82)

        // Pulsing dot (driven by draw_pass.time)
        let t = self.draw_pass.time
        let pulse = 0.5 + 0.5 * sin(t * 4.0)
        let dot_r = 4.0 + pulse * 3.0
        let dot_cx = r + 2.0
        let d_dot = length(vec2(px - dot_cx, py - cy)) - dot_r
        let dot_alpha = (1.0 - smoothstep(-1.0, 1.0, d_dot)) * bg_alpha
        let dot_color = mix(vec3(0.3, 0.6, 1.0), vec3(0.2, 0.9, 0.5), pulse)

        // Composite
        let final_rgb = mix(bg.xyz, dot_color, dot_alpha * 0.8)
        let final_a = bg.w + dot_alpha * 0.6 * (1.0 - bg.w)
        return Pal.premul(vec4(final_rgb, final_a))
    }
}

IMPORTANT: Must call self.ui.widget(cx, ids!(my_window)).redraw(cx) from handle_next_frame to keep draw_pass.time advancing. Without continuous redraw, time-based animation freezes.

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 Makepad 2.0 Shaders AI skill do?

CRITICAL: Use for Makepad 2.0 shader system. Triggers on: makepad shader, Sdf2d, pixel shader, draw_bg, draw_text, draw_quad, makepad gpu, shader function, pixel fn, vertex fn, instance, uniform, shader variable, sdf, premultiply, Pal.premul, GaussShadow, makepad graphics, custom draw, DrawQuad, DrawVector, 着色器, 像素, 渲染, 自定义绘制, 距离场

Why use Makepad 2.0 Shaders on TypingMind?

Because you install it once and use it with any model. Makepad 2.0 Shaders 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 Makepad 2.0 Shaders in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/ZhangHanDong/makepad-skills/tree/main/skills/makepad-2.0-shaders. 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 Makepad 2.0 Shaders?

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 Makepad 2.0 Shaders?

As many as you like. As long as a model supports skills, you can use Makepad 2.0 Shaders with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Makepad 2.0 Shaders AI skill free?

It is published on GitHub by ZhangHanDong. Check the repository for licensing terms. 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 👇