direct

Generates a deterministic, seedable moodboard as a single self-contained HTML file from a creative brief.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Turning a look-and-feel brief into a concrete visual direction usually requires manual moodboard assembly and produces results that cannot be reproduced or iterated on consistently. This Skill builds a deterministic, seed-driven moodboard in one HTML file so every direction can be regenerated, browsed, and verified. ## Core Features & Use Cases - Seeded generation: A seeded PRNG with named sub-streams derives palette ramps, type pairings, layout grids, and rationales from a ?seed= query parameter, guaranteeing reproducible output. - Live re-seeding UI: An input and re-seed button let viewers browse directions directly in the web-viewer pane. - Verification workflow: Census checks across a seed range, same-seed re-render diffs, and a node tools/check.mjs model validation ensure no blank, low-contrast, or clipped boards ship. - Use Case: A designer receives a brand brief asking for an editorial look; the Skill produces board/index.html with OKLCH palette ramps, contrast-noted type pairings, and a census over seeds 0-99 proving every seed renders a valid direction. ## Quick Start Create a seedable moodboard from this brief describing a brutalist editorial brand, with palette, typography, and layout all derived from the seed parameter.

Frequently Asked Questions about direct

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

FAQPage Schema
How do I create a reproducible moodboard from a design brief?

Provide a look-and-feel brief and the Skill generates a single self-contained HTML file where a seeded PRNG derives the palette, type pairing, and layout. Changing the seed query parameter produces a new direction, and the same seed always reproduces the same board.

How does seeded random generation work for design variations?

A small PRNG is derived from the hash of the seed string, with named sub-streams per concern such as palette, type, and layout. No non-seeded random calls are allowed, so every visual trait is fully determined by the seed.

Can I preview different design directions without regenerating files?

Yes, the generated board reads a `?seed=` query parameter and includes an input field plus a re-seed button. You can browse directions live in the web-viewer pane without editing or rebuilding the HTML file.

Is the moodboard output reproducible across different machines?

Same-machine reproducibility is verified by re-rendering the same seed and diffing the output. Cross-machine color consistency cannot be proven due to display and rendering differences, and the Skill states this limitation explicitly in its verdict.

How is moodboard quality verified before delivery?

Verification includes a census over the promised seed range to reject blank, low-contrast, or clipped frames, plus `node tools/check.mjs --seeds 100` which validates the pure model extracted from the artifact. Results are recorded in .harness/verdict.json.