html-wireframe is a specialist skill for the stage before styling. The agent identifies the review question, the user and the job to be done, then produces a deliberately unfinished HTML artifact that is easy to inspect, change and discuss.
When the layout is open it offers two or three structurally different directions, each with a name and a tradeoff, in a keyboard-operable selector within one file. The look stays grayscale with system type and simple blocks, so feedback focuses on structure.
What You Get
- One self-contained HTML file with inline CSS and JavaScript
- Two or three named layout directions with a one-sentence tradeoff each, or the single direction you already chose
- A grayscale, system-type look with placeholders for rich media
- Basic click-through behavior only where it helps test navigation or a short task
- A handoff listing the visual decisions deferred to a later mockup or prototype
Common Use Cases
- Comparing an overview layout and a step-by-step flow for a settings area
- Deciding where the primary action belongs on a dense screen
- Checking how a navigation model reflows from desktop to mobile
- Reviewing the content and hierarchy of a page before any branding work
What Makes It Unique
It protects the unfinished state on purpose. The skill bans brand colors, gradients, shadows and illustration so reviewers cannot drift into a styling debate, and it requires directions to be distinct product decisions. It also asks for real wording, since labels shape layout.
Under The Hood
The skill is a single instruction file with no scripts. The agent reads project and conversation context and writes one HTML file to your workspace, then opens it at desktop and mobile widths to check reading order, wrapping, overflow, focus and each click path. It needs no build tooling and no external service.