Claude Skills
CollectionsCompareWorkflowsNominate
Sign inSign up
© 2026 Curated Agent Skills·Learn more about Agent Skills
Back to repository

roast

verified

Multi-perspective UI/UX roasting workflow with iterative improvement cycles

View on GitHub

Marketplace

claude-roaster

muspelheim/claude-roaster

Plugin

claude-roaster

productivity

Repository

muspelheim/claude-roaster
1stars

skills/roast/SKILL.md

Last Verified

February 3, 2026

Install Skill

Select agents to install to:

Scope:
npx add-skill https://github.com/muspelheim/claude-roaster/blob/main/skills/roast/SKILL.md -a claude-code --skill roast

Installation paths:

Claude
.claude/skills/roast/
Powered by add-skill CLI

Instructions

# 🔥 UI/UX Roast Skill

A comprehensive skill for brutally honest UI/UX critique with multi-perspective analysis and iterative improvement.

## Core Concept

The Roast skill orchestrates multiple specialist agents to analyze UI/UX from different professional perspectives, then synthesizes their findings into actionable, brutally honest feedback.

## Execution Workflow

### Phase 1: Configuration

```mermaid
graph TD
    A[Start Roast] --> B{Target Specified?}
    B -->|No| C[Ask User for Target]
    B -->|Yes| D[Parse Target Type]
    C --> D
    D --> E{Target Type}
    E -->|Screen| F[Single Screen Mode]
    E -->|Flow| G[User Flow Mode]
    E -->|Audit| H[Full Audit Mode]
    F --> I[Configure Iterations]
    G --> J[Define Flow Steps]
    H --> K[List All Screens]
    J --> I
    K --> I
    I --> L[Detect Screenshot Method]
    L --> M[Begin Roast Loop]
```

### Phase 2: Screenshot Acquisition

**Platform Detection Order:**

1. **Xcode MCP** (iOS/macOS apps):
   ```
   Tool: mcp__xcodebuildmcp__screenshot
   Requirements: Simulator running or device connected
   Output: PNG to specified path
   ```

2. **Playwright MCP** (Web apps):
   ```
   Tool: playwright_screenshot (or equivalent)
   Requirements: Browser context active
   Output: PNG to specified path
   ```

3. **User Upload** (Fallback):
   ```
   Prompt: "Please upload a screenshot of [target]"
   Accept: File path or direct image upload
   Tool: Read (to view image)
   ```

### Phase 3: Multi-Perspective Analysis

**Parallel Agent Execution:**

| Agent | Model | Focus Area | Priority |
|-------|-------|------------|----------|
| `roaster-designer` | Sonnet | Visual design, hierarchy, color, typography | High |
| `roaster-developer` | Sonnet | Implementation feasibility, component structure | Medium |
| `roaster-user` | Sonnet | Usability, friction, task completion | High |
| `roaster-a11y` | Sonnet | Accessibility, WCAG compliance | Critical |
| `roaster-marketing` | Sonnet | Conversion, trust, brand 

Validation Details

Front Matter
Required Fields
Valid Name Format
Valid Description
Has Sections
Allowed Tools
Instruction Length:
9325 chars