メニュー
Usability heuristics for catching real problems in interfaces, not generating busywork. Nielsen's 10 and Shneiderman's Golden Rules, reframed around business outcomes with practical severity scoring.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
WCAG 2.1/2.2 guidelines, ARIA patterns, and inclusive design principles. Use when auditing accessibility, implementing ARIA, reviewing color contrast, or designing for screen readers, keyboard users, and cognitive accessibility.
Opinionated design critique framework grounded in business outcomes, named principles, and specific actionable feedback. Use when reviewing designs, running critique sessions, or giving feedback that actually improves the work.
Deceptive patterns (dark patterns), ethical design frameworks, persuasive vs. manipulative design, GDPR/consent UX. Use when reviewing designs for manipulation, designing consent flows, or evaluating ethical implications of design decisions.
Design process philosophy and frameworks. When to use JTBD, Design Sprints, Double Diamond, IDEO. How to run outcome-driven design, effective critique, and recognize when process is theater. Use when planning design work, structuring discovery, or choosing the right level of rigor for the problem.
Design systems architecture - tokens, components, documentation, governance. Use when building or scaling a design system, defining component APIs, establishing token architecture, or creating contribution guidelines.
Figma to code with AI - MCP server workflow, vibe coding, Code Connect, file preparation, design tokens, and plugin ecosystem. Use when going from Figma designs to production code with Claude Code, Cursor, or other AI coding tools.
| name | design-heuristics |
| description | Usability heuristics for catching real problems in interfaces, not generating busywork. Nielsen's 10 and Shneiderman's Golden Rules, reframed around business outcomes with practical severity scoring. |
Auto-loaded by agents:
Use when you need:
Nielsen's heuristics are from 1994. They still hold up. But the way most teams use them is broken.
The standard playbook says: get 5 evaluators, have each walk through independently, compile findings into a spreadsheet, then prioritize. What actually happens: you produce a 200-row spreadsheet that nobody reads, nothing gets fixed, and the team resents the process.
Here's what works instead: one experienced designer walks through the critical flows with these heuristics as a mental checklist. Write down the top 5 problems. Fix those. That's it. You'll catch 80% of the real issues in a fraction of the time.
Save the formal 5-evaluator method for when you're at serious scale and the stakes justify it.
The design keeps users informed about what's happening, with appropriate feedback within a reasonable time.
Why it matters: This isn't about loading spinners. It's about trust. Users who don't know what's happening leave. That's a conversion problem.
Check for: Loading indicators, progress bars, confirmation messages, state changes, active/selected states, sync status, save indicators.
The design speaks the users' language, not your team's internal jargon.
Why it matters: Every label that confuses a user is a support ticket, a bounce, or a lost sale. If your checkout says "Fulfill Order" instead of "Place Order," you're losing revenue to vocabulary.
Check for: User-facing labels, error messages, navigation labels, onboarding copy, help text. Do they use the user's vocabulary?
Users make mistakes. They need a clear way out without going through an extended process.
Why it matters: Users who feel trapped don't come back. Undo and easy exits reduce anxiety, which increases willingness to explore, which increases engagement and feature adoption.
Check for: Undo/redo, cancel buttons, back navigation, close/dismiss, exit flows, draft saving, "are you sure?" for destructive actions (but not for everything).
Users shouldn't have to wonder whether different words, situations, or actions mean the same thing.
Why it matters: Inconsistency creates cognitive load. Cognitive load slows users down. Slower users convert less. Every inconsistency is a tiny tax on every interaction.
Check for: Consistent terminology, consistent interaction patterns, consistent visual treatment, platform conventions (iOS vs. Android vs. web).
The best error message is the one that never shows up.
Why it matters: Every error is a moment where a user might quit. Preventing errors keeps users in flow, and users in flow complete tasks. Completed tasks are conversions.
Check for: Constraints (disabled buttons for invalid states), confirmations for destructive actions, smart defaults, input formatting, inline validation.
Make elements, actions, and options visible. Don't make users memorize things across screens.
Why it matters: If users have to remember something from three screens ago, some percentage of them won't. They'll abandon. This is especially brutal in mobile where context switching is constant.
Check for: Visible options vs. hidden menus, breadcrumbs, recently used items, search history, contextual help, visible labels (not just icons).
Let experienced users move fast. Shortcuts and customization for power users, simple defaults for everyone else.
Why it matters: Your power users are your highest-LTV users. If you slow them down to simplify for beginners, you'll lose the people who matter most to retention. Serve both.
Check for: Keyboard shortcuts, customization, saved preferences, recently used, favorites, bulk actions, power user features.
Every extra element on the screen competes with the things that actually matter.
Why it matters: Visual noise kills clarity. Clarity drives action. If your conversion page has 15 competing elements, users don't know where to look, and they leave. This is where editorial thinking matters. Curate, don't accumulate.
Check for: Visual noise, decorative elements competing with content, information density, whitespace usage, progressive disclosure.
Error messages in plain language. Specific problem. Clear solution.
Why it matters: A user who hits an error and can recover stays. A user who hits "Error 403" leaves and doesn't come back. Good error recovery is a retention mechanism.
Check for: Error message clarity, actionable recovery paths, specific vs. generic errors, error placement (inline vs. toast vs. modal).
The system ideally doesn't need explanation. But when it does, help should be easy to find and focused on the user's task.
Why it matters: Users who can't figure out your product don't file complaints. They just leave. Good contextual help is an activation tool, especially for complex products.
Check for: Contextual help, tooltips, onboarding flows, documentation accessibility, searchable help, empty state guidance.
| Severity | Description | Priority |
|---|---|---|
| 0 | Not a usability problem at all | - |
| 1 | Cosmetic problem, fix if time permits | Low |
| 2 | Minor usability problem, fix is low priority | Medium |
| 3 | Major usability problem, important to fix, high priority | High |
| 4 | Usability catastrophe, must fix before release | Critical |
If you have more than 3 severity-4 issues, stop building new features and fix what's broken. Shipping new features on top of a broken foundation just means more users hit more problems. You're scaling damage.
These overlap heavily with Nielsen's heuristics. Here's where they actually add something different:
The real additions from Shneiderman: universal usability (#2) and dialog closure (#4). Everything else you're already covering with Nielsen's list.
For most teams (under 50 people, shipping fast):
For teams at scale (dedicated UX research function):
The goal is to find and fix problems, not to produce a deliverable. If your heuristic evaluation produces a beautiful spreadsheet that nobody acts on, you wasted everyone's time.
laws-of-ux - Psychological principles underlying these heuristicsaccessibility-wcag - WCAG guidelines (overlaps with heuristic #2, #6, #9)