gen-art

Generates deterministic, seedable generative artwork as a single self-contained HTML file.

465|41|Updated Aug 4, 2026
One-click install
npx skills add https://github.com/autonomous-ai/openharness --skill gen-art-autonomous-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gen-art
Source: https://github.com/autonomous-ai/openharness/tree/main/store/agents/generative-art/skills/gen-art
Command: npx skills add https://github.com/autonomous-ai/openharness --skill gen-art-autonomous-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Creating generative art that is reproducible across seeds and resolutions is hard: naive sketches use unseeded randomness, break at print sizes, and cannot be browsed as editions. This Skill turns a plain-English description into a deterministic, seedable artwork in one HTML file. ## Core Features & Use Cases - Seeded determinism: Builds on mulberry32/sfc32 PRNGs with named sub-streams for composition, palette, and detail, so every seed reproduces exactly. - Seed browsing and export: Reads a ?seed= query param with a re-seed UI, renders at any size from 400px preview to 4000px print, and supports high-resolution PNG export. - Verification workflow: Runs node tools/check.mjs --seeds 100 to census seeds, checks domain invariants, and writes .harness/verdict.json with readiness status. - Use Case: Ask for a flow-field edition of 100 seeds; the Skill produces sketch/index.html with trait and rarity tables, verifies no blank or duplicate frames across seeds 0-99, and reports reproducibility guarantees. ## Quick Start Create a seedable generative flow-field artwork with an OKLCH palette and a re-seed control, then verify seeds 0 through 99.

Frequently Asked Questions about gen-art

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make a deterministic generative art sketch in HTML?

Derive a seeded PRNG such as mulberry32 or sfc32 from the hash of the seed string and never call unseeded randomness. Define the composition in logical units so the same seed renders identically at 400px preview and 4000px print.

How to create a seedable generative art edition with rarity?

Map each seed deterministically to traits like palette, layout, and density using named PRNG sub-streams per concern. Then run a census over the promised seed range, for example 0-99, to reject blank, blown-out, or duplicate frames.

Can WebGL generative art be reproducible across machines?

Same-machine reproducibility can be verified by re-rendering a seed and diffing the output. Cross-machine reproducibility for WebGL cannot be proven due to GPU and driver differences, so the verdict should state this limitation explicitly.

How do I verify a generative art seed range before publishing?

Run node tools/check.mjs --seeds 100 in the workspace, which reads the pure model from the script tag with id harness-model, checks domain invariants, and repeats each seed. Follow with browser interaction, exported-output inspection, and visual review.

Why does my generative art change when I tune the palette?

A single shared random stream reshuffles everything when one concern changes. Use named sub-streams, one per concern such as composition, palette, and detail, so tuning one aspect does not alter the others.