Ar Init

Step 3: Ask for Research Track

Ask the user to select a research track:

What type of research paper would you like to produce?

  A. Systematic literature review — evidence mapping, conceptual synthesis, gap analysis
  B. Policy or discourse analysis — policy documents, media discourse, institutional statements
  C. Comparative case study — cross-case comparison, 2-6 cases, regional or institutional analysis

Enter A, B, or C (or press Enter for auto-detect):

Map the answer:

• A → literature_review
• B → policy_analysis
• C → comparative_case_study
• Empty / other → null (auto-detect during scoping)

Step 4: Generate Project ID

.venv/bin/python3 utils/schemas.py project_id

Store the output as {PROJECT_ID}.

Step 5: Create Workspace

AR_ROOT=$(pwd)
mkdir -p "$AR_ROOT/workspaces/{PROJECT_ID}"/sources/{academic,policy,media,audit}
mkdir -p "$AR_ROOT/workspaces/{PROJECT_ID}"/analysis/{scoping,literature_map,method_cards,evidence/reading_notes,claims,debate_logs}
mkdir -p "$AR_ROOT/workspaces/{PROJECT_ID}"/drafts/sections
mkdir -p "$AR_ROOT/workspaces/{PROJECT_ID}"/reviews/{comments,revision_plans}
mkdir -p "$AR_ROOT/workspaces/{PROJECT_ID}"/{final,exports}

Step 6: Write status.json

Write workspaces/{PROJECT_ID}/status.json:

{
  "project_id": "{PROJECT_ID}",
  "topic_seed": "{USER_INPUT}",
  "initial_user_query": "{USER_INPUT}",
  "project_bootstrap_input": "{USER_INPUT}",
  "paper_type": "{TRACK or null}",
  "state": "TOPIC_RECEIVED",
  "created_at": "<ISO timestamp>",
  "updated_at": "<ISO timestamp>",
  "stages_completed": ["bootstrap"],
  "current_stage": "scoping",
  "errors": []
}

Step 7: Write config.yaml

Write workspaces/{PROJECT_ID}/config.yaml:

project_id: {PROJECT_ID}
topic_seed: "{USER_INPUT}"
paper_type: "{TRACK or auto}"
created_at: "{ISO timestamp}"

Step 8: Display Summary

Research project initialized!

  Project:    {PROJECT_ID}
  Topic:      {topic_seed}
  Track:      {track name or "auto-detect"}
  Workspace:  workspaces/{PROJECT_ID}/

Starting research pipeline...

Step 9: Auto-Launch Pipeline

Do NOT stop after creating the workspace. Immediately continue with the pipeline inline:

1. Update status.json: set state to SCOPING_IN_PROGRESS
2. Dispatch ar-scoping via the Agent tool (foreground):
   • Read and follow .claude/skills/ar-scoping/SKILL.md for workspace workspaces/{PROJECT_ID}
3. After scoping completes, read {workspace}/analysis/scoping/scoping_summary.md
4. Present GATE 1 — Topic & Scope Approval:

## GATE 1: Topic & Scope Approval

{contents of scoping_summary.md}

---

Do you approve this scope? Type one of:
- **approve** — proceed to next stages
- **revise** — tell me what to change
- **reject** — stop the pipeline

5. On approve: update status.json → state: SCOPING_APPROVED, append scoping to stages_completed. Then hand off to the orchestrator by reading and following .claude/skills/ar-orchestrator/SKILL.md for workspace workspaces/{PROJECT_ID}, starting from Stage 2 (Corpus Building).
6. On revise: re-run ar-scoping with user feedback, re-present Gate 1
7. On reject: set state to TOPIC_RECEIVED, stop

The orchestrator handles the full pipeline after Gate 1:

• Stage 2: Corpus Building — ar-retrieval + ar-source-auditor → Gate 2
• Stage 3: Framing — ar-lit-synthesis + ar-theory + ar-method → Gate 3
• Stage 4: Evidence — ar-reader + ar-evidence
• Stage 5: Argument — ar-claim-builder → Gate 4
• Stage 6: Drafting — ar-research-writer (from claims) → ar-ref-checker → ar-citation-verifier
• Stage 7: Review — ar-paper-reviewer + revision loop
• Stage 8: Export — ar-pdf-exporter → Gate 5

Guardrails

• Do NOT assume or expand the topic before the user states it
• Do NOT ask multiple nested questions before receiving the first answer
• Do NOT stop after workspace creation — always continue through the full pipeline
• The init flow is: collect topic → select track → create workspace → scoping → Gate 1 → write paper → review → export