跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
设计与 UI

ux-usability-foundations

Use when designing, critiquing, or improving an interface for usability, including affordances, feedback, constraints, error prevention, navigation clarity, or task flow.

hueyexe · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add hueyexe/frontend-agent-skills --skill ux-usability-foundations

技能说明


name: ux-usability-foundations description: "Use when designing, critiquing, or improving an interface for usability, including affordances, feedback, constraints, error prevention, navigation clarity, or task flow." license: MIT metadata: author: hueyexe

UX Usability Foundations

Purpose

Help an agent make interfaces understandable, usable, forgiving, and aligned with user expectations. Treat usability as product behavior, not surface decoration. The agent should turn unclear or fragile UI into an interface where users can quickly answer: what is this, what can I do here, where am I, what happened, how do I recover, and is this worth the effort?

When to use this skill

Use this skill when the user asks you to critique, redesign, create, or specify a UI, flow, form, app screen, website, dashboard, onboarding, settings page, checkout, navigation model, state behavior, or component interaction for usability. Use it for affordances/signifiers, feedback, constraints, errors, empty/loading/success states, recognition over recall, task clarity, and frontend behavior.

When not to use this skill

Do not use as the primary skill for deep research operations, brand styling systems, visual polish, typography craft, frontend architecture, or persuasion/growth tactics unless those choices directly affect user understanding, control, or recovery.

Core principles

  1. Start with goals, not screens. Identify the user’s goal, task, context, and likely knowledge before changing controls or layout. Design requirements describe needs before widgets.
  2. Make the next action obvious. Prefer self-evident interfaces. If the task is inherently novel or complex, make it self-explanatory with structure, labels, examples, and feedback.
  3. Show what is possible. Interactive elements must look interactive. Read-only elements must not look editable. Hidden gestures need visible alternatives.
  4. Match the user’s mental model. Use the concepts, names, and sequence users understand; do not expose database tables, file paths, internal IDs, or arbitrary system states unless users need them.
  5. Favor recognition over recall. Keep needed options, context, examples, and state visible or easy to retrieve.
  6. Reduce work and excise. Remove tasks that serve the system rather than the user’s goal: retyping, transferring data, hunting for controls, configuring obvious defaults, or answering inferable questions.
  7. Guide choices without overloading users. Use progressive disclosure, sensible defaults, grouping, and prioritization.
  8. Give immediate, informative feedback. Every consequential action needs a response that says whether the system received it, what changed, whether waiting is required, and what can happen next.
  9. Prevent errors before explaining them. Use constraints, safe defaults, previews, tolerant input, inline validation, and undo where feasible.
  10. Blame the design, not the user. Errors reveal a mismatch in expectations or an unsupported edge case. Preserve work and guide recovery.
  11. Respect platform and convention. Use familiar patterns when they fit. Break convention only when the user’s goal clearly benefits and the new behavior can be discovered and recovered from.
  12. Accessibility is usability under pressure. Design for perceptibility, operability, simplicity, and forgiveness across varied abilities, devices, literacy, stress, and environments.

See references/principle-cards.md for each principle as a reusable card.

Default recommendations

Product goal

Default to optimizing the primary user task before secondary business asks. Override only for legal, safety, fraud, privacy, or business-critical requirements. Ask: “Which obligation must interrupt or constrain the primary task, and what is the consequence of hiding or delaying it?”

Audience

Default to motivated but distracted users with average domain knowledge and imperfect memory. Override for trained expert operators using the product frequently under stable conditions. Ask: “Is this primarily for first-time/infrequent users, frequent intermediates, or trained experts?”

Platform and input

Default to platform conventions and support pointer, touch, keyboard, and assistive technology where relevant. Ask when platform/input constraints are unknown.

Information density

Default to enough context for recognition and decision-making, while hiding rare advanced controls behind clear entry points. Override for comparison, monitoring, or expert workflows. Ask: “Do users need overview density for monitoring/comparison, or step-by-step focus for completion?”

Navigation

Default to clear location, current object/state, access to major areas, and a visible route back or onward. Override only for intentional guided flows. Ask: “Should users be free to move around, or should this flow intentionally constrain navigation?”

Labels and language

Default to plain, task-oriented labels from the user’s vocabulary. Ask before using internal, clever, legal, or domain-specific terms.

Forms

Default to persistent labels, appropriate input types, upfront constraints, inline validation where helpful, preserved data, and a clear primary action. Ask only when density/speed for experts may be more important than first-time clarity.

Error handling

Default to prevention first; if error occurs, explain what happened, why it matters, and how to fix it without losing work. Ask when error details are sensitive, regulated, or security-relevant.

Destructive actions

Default to visually secondary destructive actions until the final review/confirmation step. Prefer undo for reversible actions; use confirmation for irreversible, costly, legal, or safety actions. Ask about reversibility and harm.

Feedback and status

Default to instant receipt feedback, progress for delays, visible completion, and persistent state for long-running or conse

阅读完整来源 ↗