Loading request...
Users are experiencing issues where AI tools like Claude Code and Cursor do not accurately follow the design system in Figma, leading to hallucinated components and new styles. They want a solution that ensures production-ready code with minimal human intervention while maintaining fidelity to the design system.
We haven't found the "Holy Grail" workflow yet for our UX/Product/IT sync. Our setup is pretty standard: 5 Squads, 6 Designers, and a ton of devs. We are all using **Claude (and Claude Code), Gemini Pro, Cursor, and VS Code**. Our Design System is fully built in **Figma** and mirrored in **React Native** (Mobile). However, we’re hitting a wall: Even using the **Figma MCP**, providing detailed `.md` guidelines, and setting up specific "skills" for the AI to follow, it still happens—the AI "hallucinates" components or creates new styles instead of strictly sticking to our library. **The Goal: Production-ready code with minimal hand-off.** We want to reach a point where a Figma screen (or an alternative like Pencil.dev/Paper) can be converted to production code with as few "human hands" as possible, while maintaining 100% fidelity to our DS. **My questions for the community:** * Has anyone achieved a **10/10 automated workflow**? * Are you using a specific **Context/Prompting strategy** in Cursor that actually respects a React Native library? * Should we ditch Figma for an AI-native design tool to make the bridge to Cursor/Claude Code seamless? * Is anyone using **Storybook** or **Supernova** as the "source of truth" for the AI instead of Figma? We are willing to pivot our entire toolstack if it means finding a flow that actually works and scales across 5 squads. **What is your "Golden Flow"? Thanks in advance!**