| name | pf-ai-guide |
| description | Apply Red Hat's AI design language to AI-powered features โ chatbots, assistants, generation UIs. Use when building AI experiences that should follow Red Hat brand and UX patterns. |
Red Hat AI Experience Design Patterns
This skill applies Red Hat's official design language for AI-enabled features (last updated June 2026). When building AI features, proactively apply these patterns to create transparent, polished, brand-compliant experiences.
Authoritative source: https://github.com/project-felt/ai-guidelines
Supporting documentation:
The skill references content from the project-felt/ai-guidelines repository:
- Content files:
content/*.md โ Detailed guidelines organized by topic
- Images:
assets/images/ โ Reference images showing correct and incorrect patterns
- Source:
source/ai-guidelines.txt and source/ai-guidelines.pdf โ Original design deck
Key guideline files:
content/ai-design-principles.md โ Core principles
content/transparency-notices.md โ Transparency requirements and disclosure patterns
content/iconography.md โ AI sparkle icons, usage guidelines, dos/don'ts
content/color.md โ Color usage and gradient restrictions
content/chatbot-avatars.md โ Chatbot robot avatar and launch button patterns
content/animation.md โ Animation guidelines
Core Principles
Every AI experience must follow these three principles (from content/ai-design-principles.md):
- Be transparent โ Users want to know when they are interacting with AI. Make it clear using labels and visual cues.
- Make it personable, but not human โ AI should be polite and follow Red Hat voice and tone, but shouldn't act as though the user is interacting with a human.
- Stay within the Red Hat brand and design language โ Follow PatternFly and design system standards so AI experiences look and feel like Red Hat.
Transparency Requirements
Critical: It should ALWAYS be clear when and how AI is being used (from content/transparency-notices.md).
- Don't rely on just one indicator. Use at minimum one visual AND one verbal indicator.
- Visual: Icons with AI sparkles, animations
- Verbal: "with AI", "AI-assisted", "AI-generated" in labels or button text
- High-risk interactions: Consider additional indicators (consult with AIA Reviewers)
For AI-assisted features (search, generation, editing): Place a transparency notice at the beginning of the experience. Minimum text: "This feature uses AI technology. Do not include any personal information or other sensitive information in your input." Include persistent notice: "Always review AI-generated content prior to use."
For virtual assistants/chatbots: Show notice before user interacts and before any content is generated. Include an info icon (โน๏ธ) with persistent "Always review AI-generated content prior to use." Use AI icon + "AI" tag as visual and verbal indicators.
For AI-generated content: Must include a label AND icon indicating content was created using AI.
Reference images:
assets/images/transparency-components.png โ Transparency component patterns
assets/images/info-inline-alert-ai-related-features.png โ Inline alert info variant
assets/images/warning-inline-alert-ai-related-features.png โ Inline alert warning variant
assets/images/info-modal-alert-ai-related-features.png โ Toast alert info variant
assets/images/warning-modal-alert-ai-related-features.png โ Toast alert warning variant
assets/images/virtual-assistant-ai-icon-label.png โ Virtual assistant AI indicators
Iconography Summary
Red Hat uses 9 official rh-ui-icon-ai-* icons based on sparkles (from content/iconography.md). Always pair icons with text ("...with AI", "...by AI"). Never create custom AI icons โ request via #help-brand.
Information icons:
rh-ui-icon-ai-info โ Information presented was generated by AI
rh-ui-icon-ai-error โ Problem identified by/with AI
rh-ui-icon-ai-experience โ Generic AI experience (use when others don't fit)
Action icons:
rh-ui-icon-ai-edit โ Edit something with help from AI
rh-ui-icon-ai-troubleshoot โ Receive AI help/suggestions for troubleshooting
rh-ui-icon-ai-filter โ Filter data with AI help
rh-ui-icon-ai-search โ Search with AI help
rh-ui-icon-ai-create โ Create something new with AI
rh-ui-icon-ai-enhance โ Enhance something with AI
Sparkle design rules:
- Sparkles are square with slightly round ends
- Position: upper left corner of each icon
- Sparkle breaks into icon outline with minimum 2px clear space
- Stroke ends are rounded
Icon SVGs available in repository:
All 9 official AI icons are available as SVG files in assets/images/rh-ui-icon-ai-*.svg
Reference images:
assets/images/ui-icons-ai-foundation.svg โ How sparkles are added to base icons
assets/images/sparkles-square-with-round-ends.svg โ Sparkle design details
assets/images/added-sparkle-sits-in-upper-left-corner.svg โ Sparkle positioning
assets/images/sparkle-breaks-outline-icon.svg โ 2px clearspace rule
assets/images/icons-in-use-buttons-labels-tags.png โ Icons used in UI components
assets/images/do-not-create-new-icons-with-sparkle.png โ Don't create custom icons
assets/images/do-not-use-icons-with-sparkle-without-ai-disclosure.png โ Must include disclosure
assets/images/do-not-add-sparkle-to-tech-icon.png โ Don't add sparkles to tech icons
assets/images/do-not-use-ai-sparkle-icon-for-new.png โ Don't use for "new" indicators
Chatbot Summary
All chatbots must use Red Hat's robot icon as their avatar (from content/chatbot-avatars.md). Use PatternFly non-status color tokens for avatar colors. No gradients on launch buttons or chat message boxes.
Avatar guidelines:
- Robot icon creates cohesive experience across Red Hat chatbots
- Signals advanced AI capabilities (validated by user research)
- Use PatternFly color tokens for avatar variations
- Don't use robot variations to indicate state changes
- Don't use other robot images or emoji as chatbot avatars
- Don't use robot icons for non-AI experiences or to represent humans
Reference images:
assets/images/chatbot-avatar-robot.png โ Robot icon standard
assets/images/patternfly-chatbot-ui-with-avatars.png โ PatternFly chatbot UI
assets/images/avatar-color-variants-light-theme.png โ Light theme color variants
assets/images/avatar-color-variants-dark-theme.png โ Dark theme color variants
assets/images/chatbot-launch-buttons.png โ Launch button patterns
assets/images/chatbot-launch-button-variants.png โ Launch button variants
assets/images/do-not-use-other-robot-images-as-chatbot-avatars.png โ Don't use other robots
assets/images/do-not-use-robot-variations-for-state-changes.png โ Don't vary by state
assets/images/do-not-use-robot-icons-for-humans.png โ Don't use for human users
assets/images/chatbot-donts-gradients.png โ Don't use gradients on chat UI
Color & Animation Summary
Don't use color coding or gradients to indicate AI (from content/color.md and content/animation.md). AI features use the same colors as other interface elements. Use premade sparkle animations only โ triggered on hover/click, loop once.
Color guidelines:
- Use same colors for AI as rest of UI
- Don't use gradients on AI labels/badges
- Don't use gradients on chat buttons/launchers
- Don't use gradients on chat message/composer boxes
- Don't use gradients to indicate AI "thinking" or progress
- Don't color-code AI or vary styling from rest of UI
- Don't change colors based on chatbot emotion/state
Reference images:
assets/images/do-use-same-colors-for-ai-as-rest-of-ui.png โ Correct color usage
assets/images/do-use-same-colors-for-ai-as-rest-of-ui-1.png โ More examples
assets/images/do-not-use-gradients-for-ai.png โ Don't use gradients
assets/images/do-not-color-code-ai-or-vary-from-rest-of-ui.png โ Don't vary styling
assets/images/do-not-change-styling-or-use-gradients.png โ Don't change styling
assets/images/do-not-change-color-based-on-emotion.png โ Don't use emotion colors
assets/images/chat-message-dont-gradients.png โ Don't use gradients in chat
Review Workflow
When analyzing a design against Red Hat AI patterns:
Step 1: Identify what's being used
- What AI capabilities are shown? (generation, search, troubleshooting, chatbot, etc.)
- Are transparency notices present?
- Which icons are being used?
- Is this a chatbot (should use robot avatar) or another AI feature?
- Are there visual + verbal indicator pairs?
Step 1a: Run a gradient sweep (required)
Before deciding compliance, inspect the full UI for gradients on AI-related surfaces:
- AI labels and badges (e.g., "AI Generated", "AI-assisted", "By AI")
- Chat launchers/buttons and any chatbot trigger controls
- Chat composer/message box borders, fills, and focus rings
- AI cards, panels, and highlighted callouts
- Any shimmer/glow treatment implying AI "thinking" or progress
If any of the above uses gradient color, mark as โ failed.
Step 2: Categorize each requirement
- โ
Compliant โ Meets the guideline
- โ Missing/Failed โ Does not meet the guideline
Step 3: Select reference images for failed items
Reference the appropriate image from assets/images/ for each failed check. Use "do-not-" images to show what NOT to do and corresponding "do-" images to show correct patterns.
Step 4: Summarize findings
Present a results table, then reference the correct image path for each โ item with a brief explanation of what needs to change.
Results table format:
| Category | Status |
|---|
| [Check item] | โ
Compliant |
| [Check item] | โ Missing/Failed |
For each โ item, reference the image path from the repository and explain what needs to change. The images are available at:
https://github.com/project-felt/ai-guidelines/blob/main/assets/images/[filename]
Quick Reference Checklist
Transparency:
- At least one visual indicator (icon with sparkle)
- At least one verbal indicator ("with AI", "AI-assisted", etc.)
- Appropriate transparency notice for feature type
- "Always review AI-generated content prior to use" notice where applicable
Iconography:
- Using official
rh-ui-icon-ai-* icons (not creating custom ones)
- Icons paired with text labels
- For chatbots: robot icon as avatar
Color & Styling:
- Using standard PatternFly/Red Hat colors
- No gradients on AI labels/badges, chat buttons/launchers, or chat message/composer boxes
- No gradients to indicate thinking/progress
- No special "AI colors"; follows existing color status associations
Brand Compliance:
- Follows PatternFly component patterns
- Personable but not human in tone
- Looks and feels like Red Hat
Additional Resources
Guidelines updated June 2026. This skill does not replace AI Assessment, Privacy Impact Assessment, or other required reviews.
Remember: The goal is transparency. Users want to know when they're interacting with AI. Over-communicate rather than under-communicate.