Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
By RBraga01
AI UI design quality enforcement: 8 skills and 5 agents for UI design teams
npx claudepluginhub rbraga01/a-team --plugin builder-designUse before shipping any UI that renders AI-generated content. Dynamic model output requires ARIA live regions, reading order, and cognitive load review that static content does not. Blocks "we'll do accessibility later" completions.
Use when designing or implementing any of the 6 core AI UI components. Each has specific patterns, pitfalls, and required sub-components that generic UI components don't address.
Use when designing first-run flows and empty states for AI features. AI onboarding has specific requirements — model capability communication, trust building, and graceful degradation when the model doesn't know — that generic onboarding patterns miss.
Use when designing how AI-generated content is rendered — streaming text, structured data, citations, code blocks, and uncertainty signals. Covers both visual rendering and the accessibility layer.
Use before writing any code for an AI feature's UI. All 7 states must be designed and documented before implementation begins. Blocks "we'll add loading states later" completions.
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, and rules evolved over 10+ months of intensive daily use
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Create new skills, improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, or benchmark skill performance with variance analysis.
Universal multi-agent infrastructure: 25 specialist agents, 17 enforced workflow skills, and a lead orchestrator
AI growth quality enforcement: 6 skills and 3 agents for growth teams
AI product quality enforcement: 8 skills and 5 agents for LLM product teams
AI product quality enforcement: 6 skills and 3 agents for product teams
Your AI assistant will design the chat input as a plain textarea, skip the streaming cursor, hardcode #3b82f6 in five components, and ship without a single designed error state.
This pack makes that impossible.
Generic UI patterns were not built for AI. A textarea has no concept of generation state. A div has no concept of streaming. Markdown rendered without a typography system is a wall of text. And hardcoded colours in AI components mean every rebrand is a multi-file migration.
Builder-design adds the design enforcement layer that AI feature UIs need — and that no general-purpose design system covers.
| Without builder-design | With builder-design |
|---|---|
| Streaming output has no cursor — looks broken | Streaming cursor required, aria-hidden from screen readers |
| Error state is a blank screen or generic toast | All 7 states designed before any component code |
| Prompt input is a fixed-height textarea | Auto-resize, Cmd+Enter, [Stop], token counter — or it doesn't ship |
#3b82f6 hardcoded in 6 components | Every colour, spacing, shadow references a token |
| No empty state — blank input, user abandons | 4+ clickable domain-specific examples, input focused on load |
| Screen reader announces every streaming token | role="log" aria-live="polite" aria-atomic="false" enforced |
| Skill | What It Enforces |
|---|---|
ai-states-required | All 7 states (Loading, Streaming, Success, Error, Partial, Uncertain, Empty) designed before code |
design-before-code | Spec first: layout diagram, states, copy, interaction map — then code |
accessible-ai-output | ARIA live regions, reading order, cognitive load for dynamic AI content |
ai-component-patterns | 6 AI-native components with required sub-components — no generic substitutes |
prompt-ux-design | Input pattern, suggestion system, feedback loop, history, constraints |
ai-output-design | Markdown rendering stack, typography hierarchy, confidence signals, progressive disclosure |
ai-onboarding-design | Capability communication, trust building, first win design — not just an empty input |
design-token-audit | 5 token categories, AI-specific tokens, dark mode verification — zero hardcoded values |
| Agent | Model | Role |
|---|---|---|
ai-ui-designer | Opus | Full AI component design from spec to token-compliant implementation |
ux-critic | Sonnet | Structured critique — BLOCK / WARN / NOTE verdicts, not suggestions |
state-designer | Sonnet | Maps all 7 states before any component code is written |
prompt-ux-designer | Sonnet | Prompt input, suggestion system, feedback loop, constraint design |
accessibility-reviewer | Sonnet | ARIA, keyboard, contrast, cognitive load — accessibility audit before ship |
Each skill has a Completion Statement Format — a literal code block that must be filled with real values. An agent cannot satisfy it with "looks good" or "all states are designed." It either has a state map at design/ai-states/<feature>.md or it doesn't. It either has a token count or it doesn't.
Example — ai-states-required completion:
States designed: 7/7
Feature: document-assistant
State map: design/ai-states/document-assistant.md ✓
Loading: skeleton animation, "Searching your documents..." ✓
Streaming: typing cursor (▋), [Stop] button, auto-scroll ✓
Success: response rendered, [Copy] [👍 👎] [Retry] ✓
Error: inline error, recovery action, input re-enabled ✓
Partial: truncation notice, [Continue] action ✓
Uncertain: amber left bar, "Low confidence — verify this" ✓
Empty: 4 example prompts, input focused ✓
Edge cases:
Stream interrupted mid-word: partial token cleared, Error state shown ✓
Empty string response: treated as Partial, not Success ✓
Submit during generation: queued, shown after current completes ✓
curl -fsSL https://raw.githubusercontent.com/RBraga01/builder-design/master/install.sh | bash
irm https://raw.githubusercontent.com/RBraga01/builder-design/master/install.ps1 | iex
git clone --filter=blob:none --sparse https://github.com/RBraga01/builder-design.git
cd builder-design
git sparse-checkout set skills .claude/agents
cp -rn skills/ /path/to/your/project/
cp -rn .claude/ /path/to/your/project/
builder-design is independent. Use it with any Claude project:
# Standalone
cp -r skills/ .claude/agents/ your-project/
# With A Team
# Installs alongside A Team's 25 agents and 18 workflow skills