Ar Scoping

Use WebSearch to understand the topic landscape. Run 3-5 searches:

1. "{topic_seed}" academic research — find existing scholarship
2. "{topic_seed}" literature review — find existing reviews
3. "{topic_seed}" policy debate OR public discourse — find non-academic context
4. "{topic_seed}" methodology research methods — find methodological precedents
5. (Optional) "{topic_seed}" {region/domain} recent — if the topic has geographic or temporal specificity

Record a brief summary of what was found (key themes, active debates, available source types).

Step 2: Interpret the Topic Seed

From the topic seed and web search results, extract:

• Domain: the academic discipline or interdisciplinary area
• Region: geographic scope (if any)
• Temporal scope: historical period or contemporary focus
• Disciplinary angle: which field's lens is most relevant
• Key actors or institutions: who or what is involved

Step 3: Propose Research Questions

Propose 3 candidate research questions, ranked by feasibility and specificity:

For each question:

• Question text: clear, specific, answerable
• Feasibility: high / medium / low — based on source availability
• Rationale: why this formulation is productive
• Recommended: true for the top pick, false for others

Quality criteria for a good research question:

• Specific enough to be answerable within a single paper
• Not so narrow that sources are unavailable
• Has a clear analytical or interpretive payoff
• Avoids purely descriptive framing unless doing a systematic review
• Connects to at least one existing scholarly conversation

Step 4: Define Scope Boundaries

Specify what is IN scope and OUT of scope:

• Included: topics, regions, time periods, source types, cases
• Excluded: related topics deliberately excluded and why
• Temporal range: specific years or periods
• Geographic scope: countries, regions, or global

Step 5: Identify Sub-Questions

List 2-4 sub-questions that the main research question naturally decomposes into. These will guide evidence building and section structure.

Step 6: Assess Feasibility

Evaluate:

• Source availability: are there enough accessible academic, policy, or media sources?
• Methodological fit: does the research track (lit review / policy analysis / comparative case) suit this question?
• Potential challenges: data gaps, access limitations, ethical considerations
• Preliminary source count: estimate how many sources the web search suggests exist

Step 7: Recommend Research Track

If the track was not set during init, recommend one:

• literature_review: if the topic is broad, interdisciplinary, or needs evidence mapping
• policy_analysis: if the topic centers on policy documents, institutional statements, or governance
• comparative_case_study: if the topic involves comparing 2-6 cases across regions or institutions

Step 8: Write Outputs

Write to {workspace}/analysis/scoping/scoping_report.json:

{
  "topic_seed": "{original user input}",
  "domain": "string",
  "region": "string or null",
  "temporal_scope": "string",
  "disciplinary_angle": "string",
  "refined_questions": [
    {
      "question": "string",
      "feasibility": "high|medium|low",
      "rationale": "string",
      "recommended": true
    }
  ],
  "scope_boundaries": {
    "included": ["string"],
    "excluded": ["string"],
    "temporal_range": "string",
    "geographic_scope": "string"
  },
  "sub_questions": ["string"],
  "open_uncertainties": ["string"],
  "recommended_track": "literature_review|policy_analysis|comparative_case_study",
  "preliminary_sources_found": 0,
  "feasibility_assessment": "string",
  "web_search_summary": "string"
}

Write to {workspace}/analysis/scoping/scoping_summary.md:

A human-readable summary (500 words max) that includes:

• The recommended research question (highlighted)
• Why this formulation was chosen
• What is in and out of scope
• Key sub-questions
• Feasibility assessment
• Recommended research track and why

This summary is presented to the user at Gate 1 for approval.

Step 9: Update Status

Update {workspace}/status.json:

• Set state to SCOPING_IN_PROGRESS
• The orchestrator will handle the gate approval and state transition to SCOPING_APPROVED

Quality Rules

• Do NOT assume a theory or method at this stage — that comes in Stage 3
• Do NOT generate claims or evidence — that comes in Stages 4-5
• Do NOT draft paper text — that comes in Stage 6
• Keep research questions specific and answerable
• If the topic is too broad, say so explicitly and propose narrower alternatives
• If sources appear scarce, flag this as a feasibility concern
• Always ground the scoping in actual web search results, not assumptions