Anthropic's algorithmic-art Skill: Reproducible p5.js Generative Art

## What is it, and what problem does it solve? Generative art (growing patterns from code-driven randomness) has two major pain points: (1) you get a beautiful output that can 'never be reproduced again' because the randomness wasn't seeded; and (2) it turns into 'endless parameter tweaking' with no design intent behind it. Anthropic's official algorithmic-art skill solves both with two moves — first it forces you to write an 'algorithmic philosophy' to establish design intent, then it ties every image to a seed using Art Blocks-style seeded randomness, achieving full reproducibility. ## Why this source is worth using This is a first-party skill from Anthropic's official `anthropics/skills` repo, and it ships a ready-to-use `templates/viewer.html` (with a fixed sidebar structure — Seed → Parameters → Actions, brand colors, Poppins/Lora typography, and seed-navigation buttons) plus `templates/generator_template.js` (a best-practices reference, not a "style menu"). The official docs specifically stress that the template is a "literal foundation," not just inspiration: copy its HTML structure and only swap out the p5.js algorithm and parameter controls. It also explicitly requires "creating original work, not copying existing artists" to avoid copyright issues. ## How to use it (key points) 1. Phase one: write the "algorithmic philosophy" (4-6 paragraphs) — describe the system's rules, tensions, and the feeling it's meant to evoke, not just its appearance. This step is what gives the piece a soul. 2. Phase two: turn the philosophy into a single-file interactive HTML page — p5.js embedded inline, no external dependencies. 3. Reproducibility is a hard rule: put `randomSeed(seed); noiseSeed(seed);` at the start of setup so the same seed always produces the same image. 4. Design parameters around the system's own tunable properties — count, scale, probability, ratio, angle, threshold — rather than stuffing in a pile of presets. 5. Build in seed navigation (previous/next/random/jump to a specific seed) plus Regenerate/Reset, so people can explore the entire design space within a single file. ## Common techniques Perlin-noise-driven flow fields, particle systems with vector forces, harmonic interference patterns, recursive branching structures, vector field visualizations. ## When to use it - You want to make "reproducible" generative art suitable for minting on-chain or systematic collecting (one piece per seed). - You want to turn a visual concept into a work that's "parameter-tunable and interactively explorable" rather than a static image. - You want a clean starter template for classic generative techniques like flow fields and particle systems. Source: anthropics/skills (by Anthropic, Apache-2.0 license) — this entry is an edited adaptation with rewritten content; see the link above for the original.

填下面的欄位，上方 prompt 會即時替換 [方括號] 內容。填好後按「複製組好的 prompt」直接丟進工具。

你是一位生成藝術（algorithmic art）創作者。請依下列兩階段流程，用 p5.js 幫我做一件『可重現、可互動』的演算法藝術作品。

# 主題與意圖
- 想要的視覺感受 / 美學方向：{{AESTHETIC}}
- 偏好的技法（可多選或讓我推薦）：{{TECHNIQUE}}（如：Perlin noise 流場 / 粒子系統與向量力 / 諧波干涉 / 遞迴分支 / 向量場視覺化）

# 階段一：演算法哲學
先寫 4-6 段 Markdown，描述這件作品背後的『計算美學運動』——它如何透過 noise 函數、粒子行為與數學關係，呈現你想要的感受。不要只描述外觀，要描述『系統的規則與張力』。

# 階段二：用程式表達
把哲學變成一個自包含、可互動的 HTML 作品（單檔、無外部相依，p5.js 以 inline 方式嵌入）。

# 硬規則
1. 可重現性（Art Blocks 式）：在 setup 開頭呼叫 `randomSeed(seed); noiseSeed(seed);`，確保同種子＝同畫面。
2. 參數設計：不要給一堆預設 preset，而是問『這個系統有哪些性質可以被調整？』——數量、尺度、機率、比例、角度、閾值——把這些開成 UI 參數。
3. 互動：作品內建種子導覽（Previous / Next / Random / Jump 到指定種子）、Regenerate、Reset 按鈕，讓使用者不必開新檔就能探索變體。
4. 原創性：創作全新的演算法美學，不得模仿或複製特定在世藝術家的作品，以免侵權。

請先做階段一：寫出這件作品的演算法哲學。