Canvas Generative Agent Skill

by AThevon

Algorithmic and generative art with Canvas 2D - particles, flow fields, noise, fractals, L-systems.

9k tokens

context cost

the whole folder, loaded on every use

files

instructions only

copies elsewhere

how many repositories repackaged it

167

stars on the repo

on the repository, not the skill itself

Install

one command, takes just this skill from the repository

npx skills add https://github.com/AThevon/genjutsu --skill canvas-generative

The instruction itself

17 sections, as written by the author

Canvas Generative

> Algorithmic and generative art with Canvas 2D.

> Concise rules here. Deep-dive and reference implementations in references/.

Canvas 2D Setup

DPR-Aware Sizing

Every canvas must be sharp on Retina/HiDPI displays. Set the buffer size to the physical pixel size, scale down with CSS.

function setupCanvas(canvas, width, height) {
  const dpr = window.devicePixelRatio || 1;
  canvas.width = width * dpr;
  canvas.height = height * dpr;
  canvas.style.width = `${width}px`;
  canvas.style.height = `${height}px`;
  const ctx = canvas.getContext('2d');
  ctx.scale(dpr, dpr);
  return ctx;
}

Resize Handler

function handleResize(canvas, ctx, draw) {
  const ro = new ResizeObserver(([entry]) => {
    const { width, height } = entry.contentRect;
    const dpr = window.devicePixelRatio || 1;
    canvas.width = width * dpr;
    canvas.height = height * dpr;
    ctx.scale(dpr, dpr);
    draw(); // re-render after resize
  });
  ro.observe(canvas.parentElement);
  return () => ro.disconnect();
}

Animation Loop (RAF)

let animId;
let prevTime = 0;

function loop(time) {
  const dt = Math.min((time - prevTime) / 1000, 0.1); // cap delta to avoid spiral of death
  prevTime = time;

  update(dt);
  render(ctx);

  animId = requestAnimationFrame(loop);
}

// Start
animId = requestAnimationFrame(loop);
// Stop
cancelAnimationFrame(animId);

Noise

| Type | Characteristics | Best For |

|---|---|---|

| Perlin | Smooth, grid-aligned bias, cheaper | Terrain, clouds, gentle organic textures |

| Simplex | No grid artifacts, better gradients, slightly costlier | Flow fields, organic motion, seamless tiling |

| Worley (Cellular) | Distance-to-nearest-point, cell-like | Voronoi patterns, caustics, cracks, cell textures |

Usage rules:

Always scale input coordinates (divide by a noiseScale factor) -- raw pixel coords produce visual noise
Use octaves (fractal Brownian motion) for detail: sum multiple noise calls at increasing frequency and decreasing amplitude
Seed your noise for reproducibility

// fBm pattern
function fbm(x, y, octaves = 4, lacunarity = 2, gain = 0.5) {
  let value = 0, amplitude = 1, frequency = 1, maxAmp = 0;
  for (let i = 0; i < octaves; i++) {
    value += amplitude * noise2D(x * frequency, y * frequency);
    maxAmp += amplitude;
    amplitude *= gain;
    frequency *= lacunarity;
  }
  return value / maxAmp; // normalize to [-1, 1]
}

Particle Systems

Pool Pattern (No GC Pressure)

Pre-allocate a fixed array. Never new or splice at runtime.

const POOL_SIZE = 10000;
const particles = new Array(POOL_SIZE);
let aliveCount = 0;

// Init pool
for (let i = 0; i < POOL_SIZE; i++) {
  particles[i] = { x: 0, y: 0, vx: 0, vy: 0, life: 0, maxLife: 0, active: false };
}

function spawn(x, y) {
  if (aliveCount >= POOL_SIZE) return;
  const p = particles[aliveCount++];
  p.x = x; p.y = y;
  p.vx = (Math.random() - 0.5) * 2;
  p.vy = (Math.random() - 0.5) * 2;
  p.life = 0; p.maxLife = 60 + Math.random() * 60;
  p.active = true;
}

function update() {
  for (let i = aliveCount - 1; i >= 0; i--) {
    const p = particles[i];
    p.x += p.vx; p.y += p.vy;
    p.life++;
    if (p.life >= p.maxLife) {
      // Swap with last alive, shrink pool
      particles[i] = particles[--aliveCount];
      particles[aliveCount] = p;
      p.active = false;
    }
  }
}

Flow Fields

Grid of angle vectors that steer particles. The classic generative recipe.

Build grid: Divide canvas into cells, compute an angle per cell (from noise)
Lookup: Particle position maps to grid cell, retrieve angle
Steer: Apply angle as velocity, accumulate over frames

const cols = Math.ceil(width / cellSize);
const rows = Math.ceil(height / cellSize);
const field = new Float32Array(cols * rows);

// Fill with noise-based angles
for (let y = 0; y < rows; y++) {
  for (let x = 0; x < cols; x++) {
    field[y * cols + x] = noise2D(x * 0.05, y * 0.05) * Math.PI * 2;
  }
}

// Particle follow
function followField(p) {
  const col = Math.floor(p.x / cellSize);
  const row = Math.floor(p.y / cellSize);
  if (col >= 0 && col < cols && row >= 0 && row < rows) {
    const angle = field[row * cols + col];
    p.vx += Math.cos(angle) * force;
    p.vy += Math.sin(angle) * force;
  }
  // Damping to prevent runaway velocity
  p.vx *= 0.98;
  p.vy *= 0.98;
}

Fractals & L-Systems

An L-system encodes recursive structure as string rewriting + turtle graphics.

| Component | Role |

|---|---|

| Axiom | Starting string (e.g. "F") |

| Rules | Production rules (e.g. "F" -> "F[+F]F[-F]F") |

| Angle | Turtle turn angle per +/- |

| Iterations | How many times to apply rules |

function lsystem(axiom, rules, iterations) {
  let current = axiom;
  for (let i = 0; i < iterations; i++) {
    current = current.split('').map(c => rules[c] || c).join('');
  }
  return current;
}

function drawLSystem(ctx, commands, len, angle) {
  const stack = [];
  for (const c of commands) {
    switch (c) {
      case 'F': ctx.lineTo(ctx._x += Math.cos(ctx._a) * len, ctx._y += Math.sin(ctx._a) * len); break;
      case '+': ctx._a += angle; break;
      case '-': ctx._a -= angle; break;
      case '[': stack.push({ x: ctx._x, y: ctx._y, a: ctx._a }); break;
      case ']': { const s = stack.pop(); ctx._x = s.x; ctx._y = s.y; ctx._a = s.a; ctx.moveTo(s.x, s.y); } break;
    }
  }
}

Double Buffer Pattern

Render to an offscreen canvas, then blit to the visible one. Eliminates flicker and enables trail effects.

const offscreen = document.createElement('canvas');
offscreen.width = canvas.width;
offscreen.height = canvas.height;
const offCtx = offscreen.getContext('2d');

function render() {
  // Draw to offscreen
  offCtx.fillStyle = 'rgba(0, 0, 0, 0.05)'; // trail fade
  offCtx.fillRect(0, 0, offscreen.width, offscreen.height);
  drawParticles(offCtx);

  // Blit to screen
  ctx.drawImage(offscreen, 0, 0);
}

Do Not

1. Never clearRect every frame for trail effects

Clearing destroys the trail. Use a semi-transparent fill instead.

// BAD -- kills trails
ctx.clearRect(0, 0, w, h);

// GOOD -- fades previous frame
ctx.fillStyle = 'rgba(0, 0, 0, 0.02)';
ctx.fillRect(0, 0, w, h);

2. Never getImageData in the animation loop

getImageData reads back from GPU -- extremely slow. Avoid unless absolutely required (e.g. one-time color sampling).

// BAD -- 60fps GPU readback
function loop() {
  const data = ctx.getImageData(0, 0, w, h); // blocks rendering pipeline
  processPixels(data);
  requestAnimationFrame(loop);
}

// GOOD -- sample once, cache
const colorMap = ctx.getImageData(0, 0, w, h);
function getColor(x, y) {
  const i = (y * w + x) * 4;
  return [colorMap.data[i], colorMap.data[i+1], colorMap.data[i+2]];
}

3. Always respect DPR for sharpness

A canvas without DPR scaling looks blurry on Retina displays. See setup section above.

// BAD
canvas.width = 800;
canvas.height = 600;

// GOOD
const dpr = window.devicePixelRatio || 1;
canvas.width = 800 * dpr;
canvas.height = 600 * dpr;
canvas.style.width = '800px';
canvas.style.height = '600px';
ctx.scale(dpr, dpr);

4. Never allocate in the hot loop

No new, no object spread, no array creation inside update() or render(). Pre-allocate everything.

// BAD
function update() {
  particles.forEach(p => {
    const force = { x: Math.cos(a), y: Math.sin(a) }; // new object every frame per particle
    p.vx += force.x;
  });
}

// GOOD
let fx = 0, fy = 0; // reuse
function update() {
  for (let i = 0; i < aliveCount; i++) {
    fx = Math.cos(a); fy = Math.sin(a);
    particles[i].vx += fx;
  }
}

Quick Reference: Loading Sub-skills

| Need | Load |

|---|---|

| Noise implementations, particle pool, flow field, L-system, attractors | references/algorithms.md |

| Timing and easing for animated transitions | ../motion-principles/SKILL.md |

| 3D generative (shaders, GPU particles) | ../threejs-r3f/SKILL.md |

Other skills for the same job

different authors, same section of the catalogue

Canvas Design

by anthropics

vendor ×13

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.

13k tokens

How to use it

Copy the folder

Take athevon/canvas-generative from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

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.