anthropics/skills · skills/algorithmic-art/SKILL.md
algorithmic-art
“Solid p5.js generative-art skill that writes a manifesto first and tells itself it's a master, often”
What it does
Turns a request for code art into two outputs: a 4-6 paragraph "algorithmic philosophy" in markdown, then one self-contained p5.js HTML page. The page is built on a bundled Anthropic-branded template with seed navigation and parameter sliders.
- When the agent uses it
- When a user asks for art made with code: generative art, algorithmic art, flow fields or particle systems.
- Who it is for
- Good for people who want seeded, tweakable p5.js sketches fast. Skip it if you need your own UI branding or a dark theme, or don't want a manifesto.
- What it costs
- 302 lines, about 5.3k tokens once the agent loads it.
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.
What it tells the agent to do
- Write an algorithmic philosophy: name a movement in 1-2 words, then 4-6 paragraphs, saved as a .md file
- Pick a subtle conceptual reference from the request to hide in the parameters and behaviour
- Read templates/viewer.html and use it as the literal starting point
- Write the seeded p5.js algorithm (randomSeed/noiseSeed) and a params object drawn from the philosophy
- Replace only the variable parts: algorithm, parameter sliders, optional color pickers
- Keep the fixed seed controls and action buttons, and ship one self-contained HTML file that loads p5.js from a CDN
Lines worth reading
Straight from the file, with why each one matters.
line 3Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
A clear trigger with the words people actually type. This is how a description should read.
line 49The philosophy MUST stress multiple times that the final algorithm should appear as though it took countless hours
A pep talk made mandatory. Lines 58-84 repeat it, so 'master-level' language fills context that could hold technique.
line 97Think like a jazz musician quoting another song through algorithmic harmony
The nicest idea in the file: hide a nod to the request in the maths so that only insiders notice it.
line 118Using system fonts or dark themes
Dark mode is banned outright. The UI must stay in Anthropic's light branding, whoever is using it.
line 24it should not constrain creative freedom.
What the user asks for is treated as a loose hint. Fine for an art toy, but surprising if you asked for something specific.
Watch out for
- Low
The generated page loads p5.js from cdnjs with no integrity hash, so the artwork depends on a third-party script at view time.
line 280<script src="https://cdnjs.cloudflare.com/ajax/libs/p5.js/1.7.0/p5.min.js"></script>
- Low
Step 0 requires templates/viewer.html, which could not be listed here. If it is missing, the agent will improvise the UI the skill forbids it to build.
line 390**templates/viewer.html**: REQUIRED STARTING POINT for all HTML artifacts.
- Low
User instructions are downgraded to a loose foundation, so specific requests may be reinterpreted rather than followed.
line 24it should not constrain creative freedom.
The scorecard
- TriggerSays when to use it8/10
Clear use first and real trigger words. The copyright sentence is fine but uses space.
- FocusOne clear job8/10
One job: seeded p5.js art in a branded viewer.
- LeanLight to load4/10
~400 lines, ~5.3k tokens, many repeated CRITICALs and craftsmanship refrains.
- ConcreteSteps you can follow6/10
Template, seed code and a sidebar spec, but no check step and some contradictions.
- SafeSafe to run9/10
No destructive or secret-touching actions. Only a CDN script in the output.
What it does well
- Seeded randomness plus prev/next/jump seed controls make every result reproducible and explorable.
- Splits the fixed template UI from the variable algorithm clearly, so outputs stay consistent.
- The philosophy examples (flow fields, Voronoi relaxation, L-systems) give concrete algorithmic starting points.
How it could be better
Drop the mandatory self-praise
Repeating 'master-level' phrases in the philosophy and the examples costs context and adds no technique.
Add- **Craft**: Describe the concrete choices that make it work (noise octaves, palette logic, density limits). Do not add praise phrases.
Fix the contradictions in outputs and layout
Line 7 asks for .js files but line 399 says not to create them. Download PNG is FIXED at line 330 but missing at 240. The skeleton says controls go below, but the template uses a sidebar.
AddOutput exactly two files: <name>-philosophy.md and <name>.html (all JS inline, no separate .js). Fixed Actions: Regenerate, Reset, Download PNG. Layout: canvas with a sidebar, as in templates/viewer.html.
Add a check before finishing
The skill never says what done looks like, so broken seeds or dead buttons go unnoticed.
AddBefore finishing, confirm: the page opens with no console errors; the same seed redraws an identical image; Prev/Next/Random/Jump, every slider, Reset and Download PNG all work.
Make branding a default, not a law
Forcing Anthropic fonts and banning dark themes is wrong for anyone outside Anthropic.
AddUse the template's styling by default. If the user asks for <their brand or a dark theme>, change only the CSS variables and keep the layout.
Kin reviews against Anthropic’s Extend Claude with skills and agentskills.io’s the Agent Skills standard. It reads SKILL.md only, so a script beside it can still do things the file does not say.