Install: npx skills add diskd-ai/ui-guidelines | skills.sh

Review and audit skill for applying the classic 1992 Human Interface Guidelines to desktop-style user interfaces.

This skill provides guidance and patterns for reviewing UIs against the 1992 Human Interface Guidelines, covering:

• Menus and keyboard equivalents
• Windows (document vs utility), scrolling, zooming, positioning
• Dialog boxes and alerts (modeless vs movable modal vs modal)
• Controls (buttons, radio/checkbox, sliders/steppers, disclosure)
• Icons and icon families
• Color usage (black-and-white-first design)
• Behaviors (mouse + keyboard conventions, selection/editing)
• Language (labels, messages, help systems, Balloon Help)
• Worldwide compatibility, universal access, collaborative computing UX

Triggers:

• Requests to "review my UI", "check accessibility", "audit design", or "review UX"
• Designing or auditing menus, windows, dialogs, alerts, controls, or icons
• Evaluating color usage, mouse/keyboard behaviors, or language/messages
• Checking localization, accessibility, or collaborative-computing UX patterns

Use cases:

• Audit a desktop-style UI for guideline compliance
• Review dialog box wording and layout for clarity
• Check that color is never the sole carrier of meaning
• Validate keyboard equivalents, Undo support, and progressive disclosure
• Ensure icons remain legible across selection/labeling states and bit depths

1. Establish context -- target environment, audience, artifacts, constraints
2. Classify surfaces -- identify which components (menus, windows, dialogs, controls, icons, color, behaviors, language) are in scope
3. Evaluate with checklist first, then go deep -- use references/checklist.md for high-impact issues, then references/conspect.md for rationale and patterns
4. Return findings -- high-priority violations, recommended fixes, verification checklist, open questions

ui-guidelines/
  SKILL.md                         # Entry point (routing + workflow)
  README.md                        # This file (overview)
  agents/
    openai.yaml                    # OpenAI agent interface config
  references/
    checklist.md                   # Practical review checklist (derived from Appendix C)
    conspect.md                    # Detailed paraphrased outline by chapter/topic

General feel:

• Product matches classic desktop look and feel
• Metaphors are appropriate and don't require users to "remember the mapping"
• Standard objects are not given new, conflicting behaviors

Accessibility and compatibility:

• Color is never the only carrier of meaning
• Sound is not the only carrier of meaning
• Design works across bit depths and display capabilities
• Black-and-white designs are created first; color is an enhancement

Dialogs and alerts:

• Modeless preferred; movable modal when user must respond
• Message wording turns "mysterious failure" into actionable explanation
• Spacing and focus indicators are correct

Icons:

• Legible across selection/labeling states
• Standard icon color set supports consistent behavior
• Too much black in B&W reduces selected-state contrast

Reviews should return:

1. High-priority violations -- things that break core look/feel, consistency, safety, or accessibility
2. Recommended fixes -- what to change, why, and any tradeoffs
3. Verification checklist -- what to re-check after changes
4. Open questions -- missing context that could change recommendations

• Full skill reference: SKILL.md
• Review checklist: references/checklist.md
• Detailed conspect: references/conspect.md

MIT