<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This project provides a reverse-engineered system prompt and skill library designed to transform Large Language Models (LLMs) into opinionated, accessibility-aware design collaborators. It aims to overcome the generic output often produced by AI design tools, offering a disciplined approach to creating high-quality, user-centric interfaces. The benefit is a more sophisticated and reliable AI design partner.

How It Works

The core of the project is a comprehensive system prompt, structured into 20 chapters, that instills a specific design philosophy. This philosophy emphasizes content and aesthetic discipline, robust visual hierarchy, adherence to WCAG accessibility standards, thoughtful interaction states, and system-level thinking. It's augmented by 14 procedural skills that LLMs can invoke for tasks ranging from initial discovery questions and wireframing to generating high-fidelity variations, extracting design tokens, and conducting audits. The approach prioritizes quality, depth, and respect for the design medium over superficial trends.

Quick Start & Requirements

To use the prompt, paste the contents of claude/system-prompt.md into the system prompt field of any LLM that supports this feature (e.g., Claude, GPT, Gemini, local models). The 14 procedural skills can be invoked by name as needed. The claude/ variant is calibrated for specific Anthropic models and assumes an HTML-output design environment; adaptations may be necessary for different platforms like Figma or code-only assistants. No specific hardware or software prerequisites beyond the LLM itself are detailed.

Highlighted Details

• Design Philosophy: Enforces principles like content discipline, purposeful visuals, accessibility (WCAG), system thinking, and quality over quantity.
• Procedural Skills: Offers 14 distinct skills categorized into Production (e.g., wireframe, make-a-prototype), System (e.g., design-system-extract), and Review (e.g., accessibility-audit, ai-slop-check).
• AI Trope Resistance: Explicitly designed to avoid common AI-generated design patterns such as aggressive gradients, excessive emoji use, and ubiquitous rounded corners.
• Model Calibration: Includes variants tailored for specific LLM generations (e.g., Anthropic frontier models), optimizing instruction following and skill invocation.

Maintenance & Community

The project welcomes contributions via Issues and Pull Requests, particularly for additional review skills or adaptations for different environments. No specific community channels (like Discord/Slack) or roadmap details are provided in the README.

Licensing & Compatibility

The project is released under the MIT License, permitting free use, modification, and distribution for any purpose, including commercial applications, without requiring attribution.

Limitations & Caveats

The prompt's effectiveness and interaction model are calibrated for specific, modern LLMs; older or different models may require prompt adjustments. The default workflow assumes an HTML-output environment, necessitating modifications for integration into platforms like Figma or code-centric workflows. The project is a prompt and skill library, relying heavily on the capabilities of the underlying LLM.