Wireframing And Prototyping

Wireframing & Prototyping

Use When

• You need to design a screen, flow, or feature and must decide how rough or finished the artifact should be — a paper sketch, a lo-fi wireframe, a mid-fi greybox, a hi-fi mockup, or a clickable/coded prototype.
• You are mapping a multi-screen flow and need a wireflow — wireframes wired together by the user's decisions and the system's responses (not just a static screen inventory).
• A stakeholder asks to "see it working" and you must decide whether a clickable prototype earns its cost, and at what fidelity.
• You want to validate structure, hierarchy, or a flow before investing in visual design, tokens, or front-end code.
• You need to hand a tested concept to high-fidelity design or to dev, with the open questions resolved at the cheapest possible fidelity.

Do Not Use When

• You are running interviews, surveys, or moderated usability tests, or synthesising research into decisions → use ux-research-and-usability-testing (this skill produces the stimulus you test; that skill runs the test). The two are designed to pair — see Workflow step 5.
• You are past structure-validation and producing the final visual design, tokens, or component specs → use group 09 (design-tokens-and-naming, component-library-architecture, design-handoff-and-dev-spec).
• You are designing one component's full state coverage (empty/error/loading) → use empty-error-and-loading-states (group 04). Wire those states into the wireflow here; spec them there.
• You are choosing typefaces, palettes, or doing pixel craft — that is premature at wireframe fidelity and is its own slop risk (see Anti-Patterns).

Required Inputs

• The goal of the artifact: what decision or risk this wireframe/prototype must resolve (structure? flow? feasibility? desirability? a specific interaction?). Fidelity follows from this.
• The user task(s) and entry points the flow must cover, ideally from research (personas, top tasks, JTBD). Without these you are decorating, not designing.
• The audience for the artifact: yourself (think-tool), the team (align), a test participant (validate), or a client/exec (sell). Audience sets the minimum viable fidelity.
• Known constraints: platform, key content/data, must-have steps, and any hard system responses (errors, async waits) that change the flow.

Workflow

1. State the artifact's job, then pick the lowest fidelity that does it. Name the one decision the artifact must resolve and choose the cheapest rung of the fidelity ladder that resolves it. Higher fidelity is not better — it is slower to make, harder to change, and biases reviewers toward polish over structure. See references/fidelity-ladder.md for what each rung is for and when it is the right one.
2. Wireframe the screens at that fidelity. Lay out content blocks, hierarchy, and controls — not colour, not final copy, not real type. Use real-ish content (counts, labels, lengths) so the layout is honest; use greyboxing and placeholder rectangles for media. Respect the doctrine's anti-slop rule even here: a wireframe still makes deliberate layout choices (grid, reading order, primary action) rather than reflexively dropping evenly-spaced cards (doctrine/design-doctrine.md §2).
3. Wire the screens into a wireflow. Connect each screen to the next via the user's decision and the system's response. Branch on the real forks: success vs error, empty vs populated, first-run vs return, happy path vs recovery. A flow that only shows the happy path hides exactly the risk you are prototyping. See examples/wireflow-example.md.
4. Add interactivity only where it buys validation. Make it clickable only if the transitions or timing are part of what you must test (can users find the path? does the wait feel broken?). A click-through of lo-fi screens tests navigation; a hi-fi coded prototype tests feasibility and feel. Do not raise fidelity to add clicks you do not need — see the fidelity-vs-interactivity matrix in references/fidelity-ladder.md.
5. Pair with research to test it. The wireframe/prototype is the stimulus; route to ux-research-and-usability-testing to run the protocol (tasks, think-aloud, success criteria), then fold findings back as edits at the same low fidelity before climbing. Iterating cheap is the whole point — every loop you spend at lo-fi is a loop you did not spend redoing pixels.
6. Promote, don't restart, when validated. When structure and flow are settled, hand off to high-fidelity design (group 09 handoff/tokens/components) with the resolved questions annotated on the wireflow, so hi-fi work starts from answers, not from a blank canvas.

Decision Rules

Capability Contract

• Must inspect requirements and preserve a clear simulation boundary; review is read-only unless artefact creation is requested.
• May create disposable prototypes, but may not present fake data/integrations as real, test with participants, or publish without authority.

Degraded Mode

• If no learning question exists, stop prototyping and define it before choosing fidelity.
• Without interactive tools, use an annotated storyboard and mark interaction claims untested. Recover a misleading prototype by exposing simulated areas, narrowing scope, and rebuilding only the fidelity needed.

Quality Standards

• Fidelity is no higher than necessary, all critical branches/states needed by the question exist, and realistic content replaces lorem ipsum where meaning matters.
• Evidence records question, fidelity rationale, simulated areas, test result, and next decision.

Anti-Patterns

• Starting in hi-fi. Opening Figma in full colour with real type for an unvalidated flow. It costs 5× to change, and reviewers critique the gradient instead of the journey.
• Fidelity theatre. Making it look finished to impress, when the artifact's job was to resolve structure. Polish on an unvalidated flow is wasted polish — and reads as the AI "looks-done" default the doctrine exists to defeat (doctrine/design-doctrine.md §0).
• Happy-path-only wireflows. No error, empty, loading, or recovery branch — so the prototype validates nothing that was actually risky.
• Lorem everywhere. Fake even-length text hides real overflow, truncation, and hierarchy problems. Use representative content lengths.
• Premature pixel craft. Choosing fonts, palettes, shadows, or exact spacing at wireframe stage. Those are downstream decisions; deciding them here both wastes effort and locks in choices before the structure that should drive them exists.
• Clickable for its own sake. Spending days wiring micro-interactions when a paper sketch would have answered the question.
• Never lowering fidelity to iterate. Treating each test round as a reason to add polish instead of going back down to change structure cheaply.

Outputs

• A fidelity decision: the chosen rung + the one sentence justifying it against the artifact's job.
• Wireframes at that fidelity (content blocks, hierarchy, primary action — greyboxed media).
• A wireflow: screens connected by user decisions and system responses, with the real branches (success/error/empty/first-run/recovery) shown.
• Optionally a clickable prototype, only where transitions/timing are part of what's being tested.
• A handoff packet: the validated wireflow + resolved/open questions, ready for group-09 hi-fi.

Examples

• See examples/wireflow-example.md — a worked, branch-complete wireflow for a real feature ("save a search and get notified"), with the fidelity choice stated and justified, the screen inventory, the decision/response wiring, the error/empty/loading branches, and the pair-with-research test plan. Not lorem — representative content throughout.

References

• references/fidelity-ladder.md — the five rungs (paper sketch → lo-fi → mid-fi greybox → hi-fi mockup → clickable/coded prototype): what each is for, when it is the right rung, what it cannot tell you, and the fidelity-vs-interactivity matrix.
• doctrine/design-doctrine.md — §0 (the "looks human-made" moat / no "looks-done" theatre) and §2 (the Anti-Slop Charter — deliberate layout choices apply even at wireframe stage).
• Pairs with ux-research-and-usability-testing (this group) — wireframe is the stimulus; that skill runs the test.
• For UI/web work that follows hi-fi promotion: doctrine/references/wcag-2.2-criteria.md and doctrine/references/web-performance-budgets-2026.md (validate target sizes, focus order, and perceived-performance waits that your wireflow's loading branches imply).