AgentSkillsCN

research-ideation

从主题或数据集中提炼出结构化的研究问题、可检验的假设以及实证研究策略。

SKILL.md
--- frontmatter
name: research-ideation
description: Generate structured research questions, testable hypotheses, and empirical strategies from a topic or dataset.
disable-model-invocation: true
argument-hint: "[topic, phenomenon, or dataset description]"
allowed-tools: ["Read", "Grep", "Glob", "Write"]

Research Ideation

Generate 3-5 research questions ordered from descriptive to causal.

Steps

  1. •

    Understand context: Read CLAUDE.md and any relevant project files.

  2. •

    Generate 3-5 research questions ordered by type:

    • •Descriptive: What are the stylized facts?
    • •Correlational: What patterns exist in the data?
    • •Causal: What is the effect of X on Y?
    • •Mechanism: Through what channel does X affect Y?
    • •Policy: What are the policy implications?
  3. •

    For each question, develop:

    • •Testable hypothesis with expected sign/magnitude
    • •Identification strategy (DiD, IV, RDD, synthetic control, matching, etc.)
    • •Data requirements with availability assessment
    • •Key assumptions and their plausibility
    • •Potential pitfalls and how to address them
    • •2-3 related papers
  4. •

    Rank questions by feasibility and contribution in a summary table.

  5. •

    Save output to quality_reports/research_ideation_[sanitized_topic].md.

Principles

  • •Be creative but grounded in what's feasible
  • •Think like a referee: what would make this convincing?
  • •Consider data availability realistically
  • •Suggest specific datasets (FRED, Census, PSID, BLS, central bank data, etc.)