Claude Code Plugins

Community-maintained marketplace

Feedback

moai-project-batch-questions

@modu-ai/moai-adk
137
0

Standardize AskUserQuestion patterns and provide reusable question templates for batch optimization

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name moai-project-batch-questions
version 1.0.0
created Wed Nov 05 2025 00:00:00 GMT+0000 (Coordinated Universal Time)
updated Wed Nov 05 2025 00:00:00 GMT+0000 (Coordinated Universal Time)
status active
description Standardize AskUserQuestion patterns and provide reusable question templates for batch optimization
keywords batch, questions, templates, optimization, ux, interaction
freedom medium
type project
tags project, questions, batch, templates, ux, optimization
allowed-tools AskUserQuestion

Project Batch Questions - Skill Guide

Skill Metadata

Field Value
Skill Name moai-project-batch-questions
Version 1.0.0 (2025-11-05)
Core Tool AskUserQuestion (Claude Code built-in)
UX Goal 60% interaction reduction through batching

What It Does

Purpose: Standardize AskUserQuestion patterns with reusable batch templates that reduce user interactions while maintaining clarity.

Key capabilities:

  • Batch Templates: Pre-designed question groups for common scenarios
  • UX Optimization: 60% interaction reduction through strategic batching
  • Multi-language Support: Templates in Korean, English, Japanese, Chinese
  • Response Validation: Built-in validation and processing patterns
  • Error Handling: Graceful handling of invalid or missing responses

Batch Design Philosophy

Traditional vs Batch Approach

Traditional: Q1 → Answer → Q2 → Answer → Q3 → Answer (3 interactions) Batch: Q1 + Q2 + Q3 → All answers at once (1 interaction, 66% reduction)

Batching Rules

Rule Description Example
Related Questions Group questions about same topic Language settings
Sequential Logic Q2 depends on Q1 answer Team mode conditional questions
Same Decision Context User thinking about same aspect GitHub + Git workflow

Core Batch Templates

Template 1: Language Selection Batch (3 questions)

Purpose: Set language preferences for project initialization Interaction Reduction: 3 turns → 1 turn (66% improvement)

const languageBatch = {
  questions: [
    {
      question: "Which language would you like to use for project initialization and documentation?",
      header: "Language",
      multiSelect: false,
      options: [
        { label: "English", description: "All dialogs and documentation in English" },
        { label: "한국어", description: "모든 대화와 문서를 한국어로" },
        { label: "日本語", description: "すべての対話と文書を日本語で" },
        { label: "中文", description: "所有对话和文档使用中文" }
      ]
    },
    {
      question: "In which language should Alfred's sub-agent prompts be written?",
      header: "Agent Prompt",
      multiSelect: false,
      options: [
        { label: "English (Global Standard)", description: "Reduces token usage by 15-20%" },
        { label: "Selected Language (Localized)", description: "Local efficiency with native language" }
      ]
    },
    {
      question: "How would you like to be called in our conversations? (max 20 chars)",
      header: "Nickname",
      multiSelect: false,
      options: [
        { label: "Enter custom nickname", description: "Type your preferred name using 'Other' option" }
      ]
    }
  ]
};

Template 2: Team Mode Settings Batch (2 questions)

Purpose: Configure team-specific GitHub and Git settings Interaction Reduction: 2 turns → 1 turn (50% improvement) Conditional: Only shown when mode: "team" detected

const teamModeBatch = {
  questions: [
    {
      question: "[Team Mode] Is 'Automatically delete head branches' enabled in your GitHub repository?",
      header: "GitHub Settings", 
      multiSelect: false,
      options: [
        { label: "Yes, already enabled", description: "PR merge 후 자동으로 원격 브랜치 삭제됨" },
        { label: "No, not enabled (Recommended)", description: "Settings → General에서 확인 필요" },
        { label: "Not sure / Need to check", description: "GitHub Settings 확인 후 다시 진행" }
      ]
    },
    {
      question: "[Team Mode] Which Git workflow should we use for SPEC documents?",
      header: "Git Workflow",
      multiSelect: false,
      options: [
        { label: "Feature Branch + PR", description: "매 SPEC마다 feature 브랜치 생성 → PR 리뷰 → develop 병합" },
        { label: "Direct Commit to Develop", description: "develop에 직접 커밋. 빠른 프로토타이핑에 최적" },
        { label: "Decide per SPEC", description: "SPEC 생성 시마다 매번 선택. 유연성 높지만 결정 필요" }
      ]
    }
  ]
};

Template 3: Report Generation Batch (1 question)

Purpose: Configure report generation with token cost awareness

const reportGenerationBatch = {
  questions: [
    {
      question: "Configure report generation:\n\n⚡ **Minimal (Recommended)**: Essential reports only (20-30 tokens)\n📊 **Enable**: Full analysis reports (50-60 tokens)\n🚫 **Disable**: No reports (0 tokens)\n\nAffects future /alfred:3-sync costs.",
      header: "Report Generation",
      multiSelect: false,
      options: [
        { label: "⚡ Minimal (Recommended)", description: "80% token reduction, faster sync" },
        { label: "Enable", description: "Complete reports, higher token usage" },
        { label: "🚫 Disable", description: "No automatic reports, zero cost" }
      ]
    }
  ]
};

Template 4: Domain Selection Batch (Multi-select)

Purpose: Select project domains and technology areas

const domainSelectionBatch = {
  questions: [
    {
      question: "Which domains and technology areas should be included in this project?",
      header: "Domains",
      multiSelect: true,
      options: [
        { label: "Backend API", description: "REST/GraphQL APIs, server-side logic, databases" },
        { label: "Frontend Web", description: "React/Vue/Angular, UI components, client-side" },
        { label: "Mobile App", description: "iOS/Android apps, React Native, Flutter" },
        { label: "DevOps/Infrastructure", description: "CI/CD, Docker, Kubernetes, cloud" },
        { label: "Data/Analytics", description: "Data processing, ML pipelines, analytics" }
      ]
    }
  ]
};

Response Processing

Validation Function

function validateBatchResponse(responses: Record<string, string>, template: string): ValidationResult {
  const errors: string[] = [];
  
  switch (template) {
    case 'language-batch':
      const validLanguages = ['ko', 'en', 'ja', 'zh'];
      if (!validLanguages.includes(responses['Language'])) {
        errors.push('Invalid language selection');
      }
      if (responses['Nickname']?.length > 20) {
        errors.push('Nickname must be 20 characters or less');
      }
      break;
  }
  
  return { isValid: errors.length === 0, errors };
}

Configuration Mapping

function mapToConfig(responses: Record<string, string>, template: string): Partial<Config> {
  switch (template) {
    case 'language-batch':
      return {
        language: {
          conversation_language: responses['Language'],
          agent_prompt_language: responses['Agent Prompt'] === 'English (Global Standard)' ? 'english' : 'localized'
        },
        user: { nickname: responses['Nickname'], selected_at: new Date().toISOString() }
      };
      
    case 'team-mode-batch':
      return {
        github: {
          auto_delete_branches: responses['GitHub Settings'] === 'Yes, already enabled',
          spec_git_workflow: mapWorkflowToCode(responses['Git Workflow']),
          checked_at: new Date().toISOString()
        }
      };
  }
}

Usage Integration

Alfred Command Integration

// In 0-project.md command
async function initializeProject() {
  // Step 1: Language selection batch
  const languageResponses = await executeBatchTemplate(LANGUAGE_BATCH_TEMPLATE);
  
  // Step 2: Check for team mode
  if (isTeamMode()) {
    const teamResponses = await executeBatchTemplate(TEAM_MODE_BATCH_TEMPLATE);
  }
  
  // Step 3: Report generation batch  
  const reportResponses = await executeBatchTemplate(REPORT_GENERATION_BATCH_TEMPLATE);
}

Performance Metrics

Interaction Reduction

Template Traditional Batch Reduction
Language Selection 3 interactions 1 interaction 66%
Team Mode Settings 2 interactions 1 interaction 50%
Domain Selection 5+ questions 1 interaction 80%+

Best Practices

✅ DO

  • Group related questions: Same decision context
  • Show total question count: "3 questions in this batch"
  • Use consistent headers: Short, descriptive (≤12 chars)
  • Include progress indicators: "Step 1 of 2"

❌ DON'T

  • Overload batches: Max 4 questions per batch
  • Mix unrelated topics: Keep thematic cohesion
  • Skip validation: Always verify responses
  • Ignore cancellation: Handle user gracefully

Quick Reference

Common Use Cases

Use Case Template Questions Interaction Reduction
Project initialization Language + Team batches 5 questions total 60%
Settings modification Targeted batches 1-3 questions 50-80%
Feature configuration Domain-specific batches 2-4 questions 75%

Integration Checklist

  • Template selected for use case
  • Response validation configured
  • Error handling implemented
  • Configuration mapping tested
  • Multi-language support if needed

End of Skill | Created 2025-11-05 | Optimized for batch interaction reduction