A reverse-engineered system prompt from Anthropic's Claude Design tool has emerged as an open-source framework that transforms any LLM into a disciplined design collaborator. The MIT licensed project rejects common AI aesthetic tropes like aggressive gradients and emoji-filled interfaces, instead enforcing a complete design philosophy through 20 structured chapters.

The system includes 14 procedural skills organized into production, system, and review categories. Production skills handle wireframing, prototyping, and visual variations, while review skills audit accessibility compliance and detect AI-slop patterns. The framework works across Claude, GPT, Gemini, and local models through simple system prompt integration.

Creator Trystan-SA calibrated the prompt specifically for Anthropic's frontier models, which now require less aggressive prompting due to improved instruction-following capabilities. The framework uses condition-based triggers rather than quota-based demands, allowing models to make reasonable decisions autonomously while documenting choices.

Design teams can drop the prompt into existing workflows for brand-aware design extraction or greenfield prototyping. The project welcomes contributions for additional review skills and platform adaptations, maintaining the same operational tone while expanding defensive design coverage against common AI-generated interface anti-patterns.

The framework represents a practical solution for developers seeking consistent, accessible design output from AI assistants without the typical generic SaaS-template results.