Visual Prompt Engine

Generate diverse, non-repetitive image prompts powered by real visual references from Dribbble.

AI agents tend to reuse the same visual patterns when writing image prompts. This skill breaks that cycle by grounding prompts in real, trending design work collected from Dribbble.

How It Works

Dribbble Scraper → Style Cards → Prompt Generator → Quality Reviewer → Final Prompt

1. Scrape trending designs from Dribbble (RSS or HTML)
2. Analyze each design into a structured style card (colors, composition, mood, textures, lighting)
3. Generate image prompts using 12+ distinct patterns to prevent repetition
4. Review prompts against history to ensure diversity

Installation

For AI agents (Codex, Claude Code, etc.):

npx skills add Abdullah4AI/visual-prompt-engine

For ClawHub:

clawhub install visual-prompt-engine

Quick Start

1. Install Dependencies (Optional)

The scraper works with Python standard library via RSS. For HTML scraping:

pip install requests beautifulsoup4

2. Collect Visual References

python3 scripts/scrape_dribbble.py --output data/references.json --count 20

3. Build Style Cards

python3 scripts/style_card.py build --input data/references.json --output data/style_cards.json

4. Ask Your Agent

Tell your AI agent: "Generate an image prompt for [your goal]" and it will use the style cards and prompt patterns to create a unique, design-informed prompt.

Skill Structure

visual-prompt-engine/
├── SKILL.md                          # Agent instructions (trigger + workflow)
├── README.md                         # This file (human documentation)
├── scripts/
│   ├── scrape_dribbble.py            # Collect designs from Dribbble
│   └── style_card.py                 # Build and manage style cards
├── references/
│   ├── prompt-patterns.md            # 12+ diverse prompt structures
│   ├── visual-vocabulary.md          # Precise design terminology
│   └── style-card-schema.md          # Style card JSON schema
└── data/                             # Created at runtime
    ├── references.json               # Raw scrape results
    ├── style_cards.json              # Processed style cards
    └── prompt_history.json           # Prompt dedup history

Scripts

scrape_dribbble.py

Usage: scrape_dribbble.py [--output PATH] [--count N] [--feed popular|recent|animated] [--method rss|html] [--append]

Options:
  --output, -o    Output JSON path (default: data/references.json)
  --count, -c     Number of designs to fetch (default: 20)
  --feed          RSS feed type: popular, recent, animated (default: popular)
  --method        Scraping method: rss (no deps) or html (needs requests+bs4)
  --append        Append to existing file instead of overwriting

style_card.py

Commands:
  build    Build style cards from references
  select   Select relevant cards for a prompt goal
  prompt   Generate AI analysis prompt for a card
  stats    Show style card statistics

Examples:
  style_card.py build --input data/references.json --output data/style_cards.json
  style_card.py select --goal "futuristic dashboard" --count 3
  style_card.py prompt --id sc_0001
  style_card.py stats

Automation

Set up a daily cron to keep references fresh:

# Daily refresh
python3 scripts/scrape_dribbble.py --output data/references.json --count 20 --append
python3 scripts/style_card.py build --input data/references.json --output data/style_cards.json --append

Compatibility

• Python: 3.9+
• Dependencies: Standard library only (optional: requests, beautifulsoup4)
• Agent tools: Any tool supporting the AgentSkill format (OpenClaw, Codex CLI, or custom)
• Platforms: macOS, Linux, Windows

License

MIT