refine-moodboard

Generates single-file HTML moodboards from use-case briefs using LLM-produced design specs.

Updated Feb 16, 2026
One-click install
npx skills add https://github.com/GQAdonis/artifact-refiner-skill --skill refine-moodboard-gqadonis
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: refine-moodboard
Source: https://github.com/GQAdonis/artifact-refiner-skill/tree/main/skills/refine-moodboard
Command: npx skills add https://github.com/GQAdonis/artifact-refiner-skill --skill refine-moodboard-gqadonis

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Creating a visual moodboard for a new brand or product concept normally requires manual design work; this Skill turns a plain-language brief into a rendered HTML moodboard with palettes, typography, motifs, and tone chips. ## Core Features & Use Cases - LLM-driven spec synthesis: The LLM produces structured JSON containing light and dark palettes, typography, motifs, and tone, validated for hex colors and required fields. - Template rendering: A Minijinja moodboard.html template renders the validated spec into a single self-contained HTML file with inline CSS. - Graceful fallback: When the inference proxy is unreachable or validation fails, it falls back to a placeholder mode using an existing brand's palette or a neutral gray scheme. - Use Case: Given a brief like "fintech dashboard for enterprise CFOs with a minimal, trustworthy aesthetic", generate a moodboard HTML showing color swatches, font specimens, and tone chips to align stakeholders before design work begins. ## Quick Start Ask the assistant to generate a moodboard HTML file for your use case, target audience, and aesthetic keywords, optionally anchored to an existing brand.

Frequently Asked Questions about refine-moodboard

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

FAQPage Schema
How do I generate a moodboard HTML from a text brief?

Provide a use case, audience, and comma-separated aesthetic keywords plus an output path. The skill prompts an LLM for a structured JSON spec with palettes and typography, then renders it through a Minijinja template into a single HTML file.

What inputs are required to create a design moodboard?

Four arguments are required: --use-case describing the product, --audience describing the target users, --aesthetic with comma-separated keywords, and --output for the HTML path. Optional flags include --brand, --mode, and --palette-mode.

Does moodboard generation work without an LLM endpoint?

Yes, it falls back to placeholder mode when the inference proxy is unreachable or validation fails. Placeholder mode uses an existing brand's palette if --brand is supplied, otherwise a neutral gray scheme with system fonts.

How does the skill handle invalid or malicious LLM output?

Responses are stripped of markdown fences, rejected if they contain prompt-injection markers like SYSTEM: or IGNORE PREVIOUS, and validated so every palette value is a hex literal and typography includes display, UI, and body entries.

What are the limitations of the generated moodboard?

Motif tiles and tone chips contain placeholder text in both modes since visual motif generation is out of scope. Output is limited to a single HTML file with inline CSS containing light and dark palettes.