ui

UI (from the waza collection) produces distinctive, production-grade user interfaces for pages, components, visual systems, and typography, and performs screenshot-driven visual polish. It is built to move past generic, default-prompt output by forcing a clear point of view: its guiding principle is that if a result could have been generated by a default prompt, it is not good enough. It triggers across languages, including Chinese gut-feel aesthetic complaints, and explicitly excludes backend logic and data pipelines.

The skill operates through an outcome contract (a usable interface or visual fix verified against the user's goal and relevant viewport states) and a mode picker that routes each request: bounded quick fixes to an existing screen, screenshot-grounded iteration when an image is the evidence to improve against, generated image assets (diagrams, covers, social cards), or building a new page/component/visual system. Before writing code it requires locking direction, reusing an existing app's component vocabulary and tokens when adding a surface to a mature product, and benchmarking against 2-3 mature products in the same category. Reference files cover design aesthetics, data-visualization dashboards (chart selection, tabular-nums alignment, row separators), durable-context handling with a redaction gate, and per-mode workflows. It also defers shippable documents (reports, decks, resumes, print) to a separate document-design system.

Target users are developers and designers who want frontend UI with intentional aesthetics, or who need a precise visual review of an existing screen. It suits building new components and pages, fixing cramped or incoherent layouts, iterating against screenshots, and producing information-dense dashboards.