mr-q526 AI Chat Interface PRO
Use this skill when the task is to design AI chat, copilots, agent workspaces, prompt consoles, and assistant surfaces.
Goal: produce a high-quality, production-ready page or interface that supports conversation flow, tool state, citations, safety, and trust.
Variant intent
- Use the rigorous version when quality, accessibility, conversion, and maintainability matter.
- Optimize for deep information architecture, strong hierarchy, explicit states, responsive behavior, accessibility, and validation checks.
- Prefer concrete UI decisions over generic advice.
Default workflow
- Define the primary user, task, and success action.
- Map the minimum page structure needed for the scenario.
- Establish hierarchy: what must be understood first, second, and third.
- Design responsive behavior for mobile and desktop.
- Add states: loading, empty, error, success, disabled, and permission where relevant.
- Run the final checks before delivery.
Design rules
- Use clear section names and user-facing copy.
- Make the primary action visually dominant and repeat it only when it helps.
- Keep navigation and secondary actions subordinate to the main task.
- Use spacing, typography, and alignment before adding decorative containers.
- Do not hide critical information behind hover-only interactions.
- Preserve keyboard access, readable contrast, and touch targets.
Pro guidance
- Build a strong visual and interaction thesis before writing components.
- Include proof, context, and state transitions when they affect trust.
- Use component variants intentionally: default, hover, active, focus, disabled, loading, error, and empty.
- Validate information architecture, accessibility, and responsive behavior explicitly.
- Avoid filler sections, ornamental cards, and copy that does not change user confidence.
Reject these failures
- Ambiguous primary action
- Important status or pricing hidden below the fold
- Dense UI with no scan order
- Decorative visuals that compete with the task
- Missing mobile behavior
Final checks
- Can the target user understand the page purpose in under five seconds?
- Is the next action clear without reading every paragraph?
- Are edge states and responsive behavior accounted for?