app_evaluator

Evaluates app UI screenshots against design specifications and outputs scored JSON reviews.

1.2k|126|Updated Mar 14, 2025
One-click install
npx skills add https://github.com/inclusionAI/AWorld --skill app-evaluator
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: app_evaluator
Source: https://github.com/inclusionAI/AWorld/tree/main/aworld-skills/app_evaluator
Command: npx skills add https://github.com/inclusionAI/AWorld --skill app-evaluator

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Assessing the visual quality of app interfaces is subjective and inconsistent across reviewers. This Skill applies a strict, expert-level design rubric to score app UI screenshots and deliver structured criticism and improvement advice.

Core Features & Use Cases

  • Rubric-Based Scoring: Grades interfaces on a 0-1 scale across six dimensions (design style, visual hierarchy, layout, color, components, typography) with four rank tiers from Fail to Exemplary.
  • Structured JSON Output: Returns per-image results with filename, score, rank, criticism, praise, and improvement advice in a strict JSON schema.
  • Design Guidance Library: Embeds positive and negative case analyses plus a UI design prompt hierarchy tree covering humanity, science, games, visualization, tools, and lifestyle categories.
  • Use Case: After generating multiple app UI mockups, submit the screenshots to receive comparative scores and concrete suggestions for improving weak designs.

Quick Start

Evaluate these app UI screenshots and give each one a score, rank, criticism, praise, and improvement advice in JSON format.

Frequently Asked Questions about app_evaluator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate app UI design quality automatically?

Submit app interface screenshots to the app_evaluator skill, which scores each image on a 0-1 scale across six design dimensions. It returns JSON with a score, rank tier, criticism, praise, and improvement advice per screenshot.

What criteria are used to score app interface designs?

Scoring covers design style, visual hierarchy, layout and rhythm, color palette, component detail, and typography. Scores map to four tiers: Exemplary (0.8-1.0), Good (0.6-0.79), Mediocre (0.40-0.59), and Fail (0-0.39).

Can I evaluate multiple app screenshots in one request?

Yes, the output format supports a results array where each element corresponds to one input image. The order of results matches the order of the input screenshots.

What output format does the app evaluation produce?

It produces a pure JSON object with a results array containing filename, score, rank, criticism, praise, and improvement_advice fields. Markdown code fences and unescaped double quotes inside strings are forbidden.

What are common reasons an app UI receives a low score?

Low scores result from generic card or grid layouts, default system controls, harsh or plain white color schemes, flat visual hierarchy, and uniform typography. The rubric penalizes interfaces lacking material texture, depth, and emotional design intent.