スキル一覧に戻る
yeheng

question-refiner

by yeheng

0🍴 0📅 2026年1月15日
GitHubで見るManusで実行

SKILL.md


name: question-refiner description: Transform raw research questions into structured, validated research prompts with automatic research type detection and output format validation. Ensures prompts are ready for research-executor with comprehensive quality checks.

Question Refiner

Overview

Transform vague research questions into structured, actionable research prompts through strategic clarifying questions with automatic research type detection and quality validation.

When to Use

  • User provides a raw, unstructured research question
  • Research scope is unclear or too broad
  • Need validated structured prompt for research-executor
  • Want to ensure prompt meets quality standards (≥8.0)

Core Approach

Progressive Questioning (2 rounds max):

  1. Round 1 (3 questions): Topic focus, output format, audience
  2. Round 2 (conditional): Scope, sources, special requirements
  3. Auto-detect research type → Select template → Generate & validate

Research Type Detection

TypeIndicatorsExample
Exploratory"what is", "overview", "landscape""What is the AI market like?"
Comparative"vs", "compare", "difference""Compare GPT-4 vs Claude"
Problem-Solving"how to", "solve", "fix""How to improve API performance"
Forecasting"future", "trend", "prediction""Future of quantum computing"
Deep Dive"technical", "architecture""How does BERT work internally"
Market Analysis"market", "industry", "competition""AI chip market analysis"

Output Structure

### RESEARCH TYPE
[auto-detected type]

### TASK
[Clear, specific research objective]

### CONTEXT/BACKGROUND
[Why this matters, who will use it]

### SPECIFIC QUESTIONS
1-7 concrete sub-questions

### KEYWORDS
[Search terms ≥5]

### CONSTRAINTS
- Timeframe: [e.g., 2020-present]
- Geography: [e.g., global]
- Source types: [academic, industry, news]

### OUTPUT FORMAT
- Type: [comprehensive_report|executive_summary|comparison_table]
- Citation style: [inline-with-url|footnotes]

### QUALITY SCORE
[0-10, must be ≥8.0]

Quality Validation

ComponentWeightCriteria
Completeness30%All required fields present
Specificity30%Questions are specific, not vague
Keyword Richness20%≥5 search terms with synonyms
Constraint Clarity20%Clear, realistic constraints

Process: Generate → Validate → If score < 8.0: Refine (max 2 attempts)

Token Optimization

📋 Reference: .claude/shared/constants/token_optimization.md

Context Budget: 10k tokens max

Error Handling

📋 Reference: .claude/shared/constants/error_codes.md

  • E001: Insufficient context → Ask clarifying questions
  • E003: Validation failed → Refine and retry
  • E004: Quality < 8.0 after retries → Request manual review

See also: Skill Base Template

Examples

See examples.md for detailed interaction patterns.

Detailed Instructions

See instructions.md for complete questioning strategy.

スコア

総合スコア

60/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

+10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です