threerocks/hand-drawn-styles

What it solves

It eliminates the tedious process of manually searching for, combining, and refining prompts to achieve specific hand-drawn artistic styles in AI image generation. Instead of guessing keywords or repeatedly tweaking descriptions, users can apply pre-verified "recipes" to ensure consistent visual results across different image models.

How it works

The project provides a tool-agnostic set of protocols and style recipes. It is designed to be installed as a "skill" or custom instruction for AI Agents (such as Claude Code, Cursor, or Gemini CLI). When a user asks to draw something in a specific style, the Agent uses the provided recipes to:

1. Select a Style: Use a specified style (by number, name, or alias) or provide a menu for the user to choose from.
2. Fill Placeholders: Automatically infer and insert the subject, actions, and text into the style's prompt template.
3. Generate the Prompt: Output a clean, final prompt that the user can then copy into an image generator like Midjourney or GPT-image.

For high-stability production, it includes a Python renderer (render_prompt.py) that extracts recipes exactly as written to prevent the AI Agent from accidentally altering the style.

Who it’s for

• AI Artists and Content Creators: Who want a wide variety of hand-drawn styles without writing complex prompts from scratch.
• Agent Users: People using AI coding assistants or LLM-based CLI tools who want to integrate image prompting capabilities directly into their workflow.
• Developers: Who need consistent, repeatable visual styles for project assets.

Highlights

• 19 Verified Styles: Includes a diverse range of styles such as Ghibli-esque, xkcd-style stickmen, Nordic storybooks, 3D vinyl toys, and "intentionally bad" crayon doodles.
• Tool Agnostic: Works across various AI Agents (Claude Code, Cursor, Windsurf, etc.) via simple markdown rules.
• Placeholder System: Automatically handles subject and layout inference so users only need to provide the core idea.
• Production-Ready: Offers a JSON output format and a specific multi-stage workflow for complex styles (like style 3.1) to ensure professional-grade consistency.