What problem does it solve? AI coding assistants produce visually identical, template-like interfaces that audiences instantly recognize as generated. Hallmark enforces structural variety, honest content, and design-token discipline so generated pages and components look deliberately designed rather than defaulted. ## Core Features & Use Cases - Design flow with macrostructure and theme rotation: Picks one of 21 named macrostructures and one of 22 catalog themes (or a custom OKLCH palette), with mandatory diversification rules so consecutive outputs never repeat structure or theme axes. - Audit, redesign, and study verbs: Scores existing code against a named anti-pattern catalog, rebuilds visual structure while preserving content and IA, or extracts design DNA from screenshots and URLs into a portable design.md. - Component-scope mode with 8-state discipline: Detects single-element briefs and emits a component plus a demo wrapper rendering all 8 interaction states (default, hover, focus, active, disabled, loading, error, success). - Use Case: Ask your AI assistant to build a landing page for a coffee subscription; Hallmark runs a pre-flight scan of your project, asks about audience and tone, picks a macrostructure and theme, and emits token-locked, mobile-verified HTML/CSS stamped with its choices. ## Quick Start Ask your AI coding assistant to design a landing page for your product and let Hallmark ask its three context questions before it builds.