This project provides a standardized, tool-agnostic library of prompt templates for generating images in various hand-drawn styles. It addresses the fragmentation of effective prompts by consolidating 14 distinct, verified styles into reusable recipes. The target audience includes users of AI agents like Claude Code, Cursor, and Gemini CLI, enabling them to quickly generate high-quality, consistent hand-drawn image prompts without manual searching or piecing together.

How It Works

The core approach leverages two tool-agnostic files, PROTOCOL.md and STYLES.md, containing prompt templates for 14 hand-drawn styles. Users describe their desired content to an AI agent, which then uses these files to select a style (or prompts the user to choose), automatically fills content placeholders (e.g., 【主体】, 【主色调】), and outputs a clean, ready-to-use prompt. This design separates prompt generation from image generation, maintaining portability and flexibility. Styles retain their native structures, and aspect ratios are handled non-rigidly, preserving artistic integrity.

Quick Start & Requirements

• Primary install/run command: git clone https://github.com/threerocks/hand-drawn-styles.git
• Integration: Include PROTOCOL.md and STYLES.md within your AI agent's rule or skill directory (e.g., .cursor/rules/, ~/.claude/skills/hand-drawn).
• Prerequisites: An AI agent capable of processing custom instructions or rules.
• Links: examples/ directory contains sample images and prompts.

Highlighted Details

• Features 14 distinct hand-drawn style recipes (12 main styles + 2 variants).
• Styles are categorized into Realistic Hand-drawn, Line Art/Explanation, Deliberately Bad Art, and Traditional/Retro.
• Supports automatic inference and filling of content placeholders within templates.
• Offers flexible aspect ratio handling, avoiding rigid constraints.

Maintenance & Community

The project welcomes contributions via Pull Requests for new, verified hand-drawn styles. Specific details on core maintainers, sponsorships, or dedicated community channels (like Discord/Slack) are not explicitly detailed in the README.

Licensing & Compatibility

The project is released under the MIT License. This permissive license allows for broad compatibility, including commercial use and integration into closed-source applications without significant restrictions.

Limitations & Caveats

This repository exclusively generates prompt text and does not perform image generation itself. The handling of aspect ratios for layout-specific styles is described as "soft structure" rather than hard numerical constraints. The effectiveness of placeholder inference is dependent on the AI agent's interpretation capabilities.