interaction-design-sop

Frontend interaction resilience & defense SOP for engineering implementation. Covers 7-step pre-coding interaction checklist: user intent & task friction reduction, P0/P1/P2 information hierarchy triage, 4-state resilience (empty, loading, error, destructive confirmation), perceived performance (optimistic UI, skeletons), power user shortcuts, and layout resilience trade-offs. Use when: implementing UI views/forms/dashboards in code, designing component interaction logic, handling edge cases and states, or conducting frontend UX resilience reviews. Keywords: 交互细节, 状态设计, 边界情况, 异常处理, 交互审查, 骨架屏, 乐观更新, 防错设计, interaction design, edge cases, error state, empty state, loading state, UI review, UI resilience.

Interaction Design SOP — Frontend Engineering Resilience Framework

A rigorous pre-coding interaction resilience & defense SOP for frontend engineers and interaction designers. Focuses on interface completeness, error resilience, and edge-case handling before writing code.

[!NOTE] System Collaboration Guidelines:

• Product UX Lifecycle: For end-to-end product design (user scenarios, business metrics, journey maps, and wireframe prototypes), refer to product-ux-workflow.
• Component Implementation: For code implementations, consult custom-ui-spec (for headless components like shadcn/ui and Radix) or ui-framework-spec (for pre-built libraries like Ant Design and Element Plus).

When designing a UI, work through these 7 steps before writing code. Each step constrains the next — decisions compound, so get the early steps right.

Step 1: Intent & Context

Define who is using this and what they're trying to accomplish. If you can't state the primary goal in one sentence, you don't understand the task yet.

• What specific scenario brings the user to this interface?
• What is the single most important action they need to complete?

Output: a clear user scenario and the one primary goal. Reject feature creep at this stage — secondary goals come later.

Step 2: Task Flow Optimization (Friction Reduction)

Map the happy path from entry to completion. Then cut it.

• List every click, decision, and input the user must make
• Eliminate steps: smart defaults, pre-fill from context, merge multi-step forms into one
• A step that can be automated should be automated

Target: the minimum number of actions to complete the primary goal. Each remaining step must justify its existence.

Step 3: Information Architecture

Triage every piece of information into three tiers:

P0 — Vital. The content users spend 80% of their time on. Gets the best screen real estate, largest visual weight, and appears without any user action. If you had only one screenful to show, this is what stays.

P1 — Secondary. Supporting context and actions. Available without hunting, but never competes with P0 for attention. Smaller, lower contrast, positioned around P0.

P2 — Progressive. Edge-case settings, advanced options, low-frequency data. Hidden behind disclosure: collapsible sections, hover tooltips, "advanced" toggles, or secondary tabs. Default: hidden.

For each piece of data on the screen, assign a tier. If you can't decide, it's P2.

Step 4: Resilience & Edge Cases

Design for three states every view must handle:

Empty state. What shows when there's no data? A blank screen is a dead end. Show a clear next action — a create button, sample data, or an inline prompt. The user should know exactly what to do next.

Error state. What shows when the network drops or the API fails? Never show a raw error. Provide a human-readable message, a retry action, and preserve any user input so they don't lose work. Consider offline fallback.

Loading state. What shows while data fetches? Skeleton screens for content areas, spinner for short actions, progress bar for uploads. Never flash a blank screen then suddenly populate — that feels broken.

Destructive actions. Identify every action that can't be undone (delete, overwrite, bulk operations). Each needs at minimum a confirmation step. For truly catastrophic actions, require the user to type the entity name to confirm.

Step 5: Perceived Performance & Power Users

Speed is a feature, not a metric. Optimize for perceived speed even when actual latency exists:

• Optimistic UI: Show the expected result immediately, then reconcile. Works for likes, toggles, quick edits. Roll back on failure.
• Skeleton screens: Show layout structure instantly while data loads. Beats a spinner — gives users something to scan.
• Prefetching: Load data the user will likely need next (hover-triggered prefetch, route-based prefetch).

For power users, add keyboard shortcuts for the top 3-5 actions. Consider whether a Cmd+K command palette is warranted for discoverability.

Step 6: Trade-off Critique

Before writing code, stress-test the design against three perspectives:

• Layout resilience: Does it break with very long text? 320px-wide viewport? Right-to-left languages? High-contrast mode? 200% font size?
• Power user experience: Can an expert drive this entirely by keyboard? Can they batch-operate? Is there a shortcut for the thing they do 50 times a day?
• Novice experience: Would a first-time user know what to do within 5 seconds? Is complexity unnecessarily exposed? Is there a guided path?

Then make a deliberate trade-off: what did you sacrifice and why? Document it. "It's simple because we dropped bulk-edit" is a decision; "it's complex because we added everything" is an abdication.

Step 7: Output

Produce the actual UI implementation. The format depends on the task:

• Standalone page/component: working React/Vue/HTML code with proper states
• System design: component tree, data flow diagram, route structure
• Design review: annotated screenshots or a written audit against the 6 steps above

Whatever the format, it must handle the empty, loading, error, and destructive-action states defined in Step 4.

Execution Rules

1. Work through all 7 steps in order. Skipping steps produces fragile designs.
2. Every step must produce a concrete artifact — not just agreement, but specific decisions.
3. P0 content gets the best real estate. P2 content starts hidden. This is non-negotiable.
4. Every view must define its empty, loading, and error states. A view without all three is incomplete.