ux-ai-automation
Skill21 starsChanged 5 months agoMITView on GitHub
AI feature and automation UX: AI assistants, smart recommendations, semantic search, predictive input, AI transparency and labeling, user control, feedback mechanisms, and workflow automation. Use when designing or reviewing any AI feature, LLM integration, personalization engine, or user-facing automation.
In this repository17 other files it gives its agents.
- ux-accessibilitySkill
- ux-contentSkill
- ux-error-handlingSkill
- ux-formsSkill
- ux-generalSkill
- ux-help-onboardingSkill
- ux-information-architectureSkill
- ux-navigationSkill
- ux-networkSkill
- ux-notificationsSkill
- ux-reviewSkill
- ux-safety-privacySkill
- ux-searchSkill
- ux-settingsSkill
- ux-user-accountSkill
- ux-utilitySkill
- ux-visual-designSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 42
Explainability: Always answer "why is the AI showing me this?" — either proactively or available on demand.
- Distinctive: line 46
Graceful degradation: When confidence is low, the AI should say less, not guess more.
- Distinctive: line 57
Transparency by default: Every AI-generated piece of content should be labeled.
- Distinctive: line 64
Use a consistent badge across the entire product (a sparkle icon is widely recognized).
- Distinctive: line 98
When applying this checklist to a screen, flow, or codebase:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI and Automation UX
AI features fail not because the model is bad, but because the experience around it is poorly designed. Users need to trust the AI, understand what it can and cannot do, stay in control, and recover gracefully when it is wrong.
Checklist
| Item | Guidance |
| AI assistant / chatbot | Clear entry point. Example prompts shown to start. Scope communicated upfront. |
| Smart recommendations | Explains why ("Because you liked X"). Includes a "Not interested" option. |
| AI-powered search | Supports natural language queries. Shows what the AI understood from the input. |
| Predictive input | Non-intrusive. Tap to accept, ignore to dismiss. Never auto-inserts without confirmation. |
| Auto-categorization | Shows the AI-assigned label and lets the user correct it inline. |
| AI-generated summaries | Clearly labeled as AI-generated. Links to the original source. |
| AI transparency | All AI-generated content is marked with a consistent icon or badge. |
| User control over AI | Per-feature opt-out in Settings. Not a single global "Disable AI" toggle. |
| AI error handling | Falls back gracefully when uncertain. Never exposes a raw model error to the user. |
| On-device AI | On-device processing is used for sensitive content (photos, notes, health data). |
| Feedback mechanism | Thumbs up / down on every AI output. Minimal friction to give feedback. |
| Automation and workflows | Preview of what the automation will do is shown before the user confirms execution. |
| Context-aware suggestions | Suggestions adapt to the user's current task, time of day, and usage history. |
| AI model update disclosure | Users are informed when the AI model powering a feature changes significantly. |
AI UX Principles
Explainability: Always answer "why is the AI showing me this?" — either proactively or available on demand. A recommendation with a visible reason is trusted. One without a reason is suspicious.
Graceful degradation: When confidence is low, the AI should say less, not guess more. "I'm not sure — here are 3 options" is better than one confident wrong answer.
Progressive trust: Earn trust over time. Start with conservative recommendations. Let users see that the system is improving with their use.
User in the loop: AI proposes, the user decides. Never execute an irreversible action without explicit user confirmation. For automation, show a preview and require a deliberate "Run" tap.
Transparency by default: Every AI-generated piece of content should be labeled. Users should always be able to ask "How was this made?" and get a clear answer.
AI Content Labeling
Use a consistent badge across the entire product (a sparkle icon is widely recognized). Place it near the content, not buried in metadata. On tap or hover, show a brief explanation: "This was generated by AI." Never mix AI-generated and human-created content in the same feed without differentiation.
Feedback Loop Design
Micro-feedback (low friction): Thumbs up / thumbs down visible on hover or long-press. "Was this helpful?" at the end of AI responses.
Macro-feedback (periodic): "How is this feature working for you?" prompt after 7 days of use. A dedicated feedback option in Settings > AI.
Where possible, show users that their feedback has an effect: "Your suggestions have been used to improve your recommendations."
On-Device vs Cloud AI
| Criteria | On-device | Cloud |
| Privacy | High — data stays on device | Lower — data sent to server |
| Latency | Very low | Depends on connection speed |
| Capability | Limited by device hardware | Full model capability |
| Cost | One-time model download | Per-inference cost |
| Best for | Photos, text suggestions, classification | Chat, generation, summarization |
How to Review
When applying this checklist to a screen, flow, or codebase:
- Go through every checklist item and mark it Pass, Fail, or N/A — never skip items silently.
- Cite specific evidence for each finding: the screen, component, file, or line it applies to.
- Tag each failure with a severity: Blocker (breaks the experience), Major (hurts usability), or Minor (polish).
- End the review with the top 3 recommended fixes, ranked by user impact.
Related Skills
ux-content— labeling and disclosure copy for AI featuresux-safety-privacy— on-device processing and data handlingux-search— AI-powered search within the broader search experience
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.