agentsclimarketplace

Roast

Skill muspelheim/claude-roaster/skills/roast

Multi-perspective UI/UX roasting workflow with iterative improvement cyclesFrom its SKILL.md

Install
npx -y skills add muspelheim/claude-roaster --skill roast

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.4 KB, as published. Nobody here has run it

πŸ”₯ UI/UX Roast Skill

Brutally honest UI/UX critique with multi-perspective analysis and iterative improvement.

Command Syntax

/roast [mode] [target] [options]

# Modes
screen <target>     # Single screen analysis (default)
flow <target>       # Multi-screen user journey
audit               # Full application review

# Options
--iterations=<1-10> # Roast cycles (default: 3)
--focus=<area>      # Priority: a11y|conversion|usability|visual|implementation
--output=<path>     # Output directory (default: reports/roast/)
--fix=<mode>        # Fix handling: auto|report|ask (default: ask)

Examples

/roast screen login                     # Roast login screen
/roast screen settings --focus=a11y     # Focus on accessibility
/roast flow checkout --iterations=5     # 5-iteration checkout flow
/roast flow onboarding --fix=auto       # Auto-fix issues found
/roast audit                            # Full app audit

Execution Workflow

1. Immediate Start

Do NOT ask questions upfront. Apply smart defaults and begin immediately.

πŸ”₯ ROAST SESSION STARTED
β”œβ”€ Mode: screen
β”œβ”€ Target: login
β”œβ”€ Iterations: 3
β”œβ”€ Focus: balanced
└─ Output: reports/roast/

2. Screenshot Capture

Auto-detect screenshot method:

πŸ“Έ Capturing screenshot...
β”œβ”€ Xcode MCP: [βœ“ found | βœ— not found]
β”œβ”€ Playwright MCP: [βœ“ found | βœ— not found]
└─ Using: [xcode | playwright | manual upload]

Detection priority:

  1. mcp__xcodebuildmcp__screenshot β†’ iOS/macOS
  2. mcp__playwright__browser_take_screenshot β†’ Web
  3. Request user upload β†’ Fallback

CRITICAL: Always specify the output path explicitly!

// First, ensure directory exists
Bash: mkdir -p reports/roast/screenshots

// Playwright MCP - specify filename parameter
mcp__playwright__browser_take_screenshot({
  type: "png",
  filename: "reports/roast/screenshots/[target]_[iteration].png"
})

// Xcode MCP - specify path parameter
mcp__xcodebuildmcp__screenshot({
  path: "reports/roast/screenshots/[target]_[iteration].png"
})

Without explicit path, screenshots go to wrong location (e.g., .playwright-mcp/).

3. Parallel Analysis with Progress

Show real-time progress:

πŸ”₯ Roasting: login (1/3)
β”œβ”€ πŸ“Έ Screenshot βœ“
β”œβ”€ 🎨 Designer: analyzing...
β”œβ”€ πŸ’» Developer: analyzing...
β”œβ”€ πŸ‘€ User: analyzing...
β”œβ”€ β™Ώ A11y: analyzing...
└─ πŸ“ˆ Marketing: analyzing...

Launch agents in parallel:

// All 5 agents run simultaneously
Task(subagent_type="claude-roaster:roaster-designer", prompt="...")
Task(subagent_type="claude-roaster:roaster-developer", prompt="...")
Task(subagent_type="claude-roaster:roaster-user", prompt="...")
Task(subagent_type="claude-roaster:roaster-a11y", prompt="...")
Task(subagent_type="claude-roaster:roaster-marketing", prompt="...")

Update as each completes:

β”œβ”€ 🎨 Designer: βœ“ 3 issues
β”œβ”€ πŸ’» Developer: βœ“ 2 issues
β”œβ”€ πŸ‘€ User: βœ“ 4 issues
β”œβ”€ β™Ώ A11y: βœ“ 5 issues (2 critical!)
└─ πŸ“ˆ Marketing: βœ“ 2 issues

4. Results Summary

Display in terminal before asking about fixes:

πŸ”₯ ROAST RESULTS (Iteration 1/3)

Found 16 issues:
β”œβ”€ πŸ”΄ Critical: 2
β”œβ”€ 🟠 Major: 6
└─ 🟑 Minor: 8

Top Critical Issues:
1. Missing form labels - add aria-label to inputs
2. Contrast ratio 2.1:1 - increase to 4.5:1 minimum

πŸ“„ Full report: reports/roast/roast_login_1.md

5. Fix Decision (After Results)

Ask ONLY after showing results:

How should we handle these 16 issues?

[1] Auto-fix critical & major (8 fixes)
[2] Fix all issues (16 fixes)
[3] Cherry-pick fixes
[4] Report only (no changes)

6. Fix Implementation

If fixes chosen:

πŸ”§ Implementing fixes...
β”œβ”€ [1/8] Adding aria-labels... βœ“
β”œβ”€ [2/8] Fixing contrast... βœ“
β”œβ”€ [3/8] Increasing touch targets... βœ“
...
└─ βœ“ Complete

πŸ“Έ Capturing updated screenshot...

7. Next Iteration

Iteration 2/3 starting...
[Repeat steps 2-6]

8. Final Summary

πŸ”₯ FINAL ROAST SUMMARY

Session Complete!
β”œβ”€ Iterations: 3
β”œβ”€ Issues found: 24
β”œβ”€ Issues fixed: 18
└─ Resolution: 75%

Score Improvement:
| Category      | Before | After | Ξ”    |
|---------------|--------|-------|------|
| Visual        | 4/10   | 8/10  | +4   |
| Usability     | 5/10   | 9/10  | +4   |
| Accessibility | 3/10   | 8/10  | +5   |
| Overall       | 4/10   | 8/10  | +4   |

πŸ“„ reports/roast/roast_login_final.md

Mode-Specific Behavior

Screen Mode

Single screen, multiple iterations:

/roast screen login --iterations=3

Iteration 1: Capture β†’ Analyze β†’ Report β†’ Fix?
Iteration 2: Capture β†’ Analyze β†’ Report β†’ Fix?
Iteration 3: Capture β†’ Analyze β†’ Final Report

Flow Mode

Multiple screens, analyze journey:

/roast flow checkout

πŸ—ΊοΈ Flow: checkout (5 steps)
β”œβ”€ [1/5] Cart β†’ πŸ“Έ analyzing...
β”œβ”€ [2/5] Shipping β†’ πŸ“Έ analyzing...
β”œβ”€ [3/5] Payment β†’ πŸ“Έ analyzing...
β”œβ”€ [4/5] Review β†’ πŸ“Έ analyzing...
└─ [5/5] Confirmation β†’ πŸ“Έ analyzing...

Cross-screen checks:
β”œβ”€ Visual consistency
β”œβ”€ Navigation clarity
β”œβ”€ Progress indication
└─ Drop-off risk points

Audit Mode

Auto-detect and roast all critical screens:

/roast audit

πŸ” Scanning for critical screens...
β”œβ”€ Login βœ“
β”œβ”€ Dashboard βœ“
β”œβ”€ Settings βœ“
β”œβ”€ Checkout βœ“
└─ Profile βœ“

Roasting 5 screens (3 iterations each)...

Agent Configuration

AgentModelFocusWeight
roaster (orchestrator)OpusSynthesis-
roaster-designerSonnetVisual, typography, color1.0x
roaster-developerSonnetImplementation, structure1.0x
roaster-userSonnetUsability, friction1.0x
roaster-a11ySonnetAccessibility, WCAG1.0x
roaster-marketingSonnetConversion, trust1.0x

With --focus option:

  • Focused agent: 1.5x weight
  • Other agents: 0.5x weight

Output Structure

reports/roast/
β”œβ”€ roast_[target]_1.md
β”œβ”€ roast_[target]_2.md
β”œβ”€ roast_[target]_final.md
└─ screenshots/
   β”œβ”€ [target]_1.png
   β”œβ”€ [target]_2.png
   └─ [target]_final.png

Report Format

# πŸ”₯ Roast Report: [Target] - Iteration [N]

**Mode:** screen | **Focus:** balanced | **Date:** 2024-01-15

![Screenshot](screenshots/login_1.png)

## The Verdict

[Brutal 2-3 sentence summary]

## Issues by Severity

### πŸ”΄ Critical (2)
| Issue | Agent | Fix |
|-------|-------|-----|
| Missing labels | A11y | Add aria-label="Email" |
| Low contrast | Designer | Change #999 to #595959 |

### 🟠 Major (6)
...

### 🟑 Minor (8)
...

## Agent Deep Dives

### 🎨 Designer
[Full analysis]

### β™Ώ A11y Expert
[Full analysis]

...

## Quick Wins
- [ ] Fix 1 (< 2 min)
- [ ] Fix 2 (< 2 min)

## Scores
| Category | Score |
|----------|-------|
| Visual | 4/10 |
| Usability | 5/10 |
| Accessibility | 3/10 |
| Overall | 4/10 |

Voice Guidelines

  1. Brutal but fair - Harsh critique, always with solutions
  2. Specific values - "#2563eb", "48px", "font-weight: 600"
  3. Actionable fixes - Every issue has a concrete fix
  4. Fast start - Begin immediately, ask questions later
  5. Progress visibility - Always show what's happening

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.