Analysis Reflection

Step 2: Present Analysis Summary

Show structured overview:

── Reflection: {design_id} ──
Title: {title}
Hypothesis: {hypothesis_statement}
Intent: {analysis_intent}
Status: {status}

── Journal Overview ──
Total events: {count}
Phase: {inferred_phase}  (see /analysis-journal Phase Inference rules)
Timeline: {first_event.created_at} → {last_event.created_at}

── Evidence ──
Supporting: {count} items
  - {content} ({created_at})
Contradicting: {count} items
  - {content} ({created_at})

── Decisions Made ──
  - {method} ({package}) — {reason}

── Open Questions ──
  - {question_content} ({event_id})

── Branches ──
  - {child_design_id}: {child_title} ({child_status})

To find branches, use:

children = [d for d in list_analysis_designs(theme_id=design.theme_id) if d.parent_id == design_id]

Step 3: Guided Reflection (3 Questions)

Ask in sequence. Record each answer as a reflect event in the journal.

Q1: Evidence Assessment "証拠を総合すると、仮説「{hypothesis_statement}」についてどう言える？"

• Supporting evidence dominant → supported direction
• Contradicting evidence dominant → rejected direction
• Inconclusive → inconclusive direction or more investigation

Q2: Open Questions Check If open questions exist: "未解決の問い ({count}件) がある。これらは結論に影響する？"

• Affects conclusion → recommend more investigation (→ /analysis-journal)
• Does not affect → proceed to conclusion

Q3: Conclusion or Next Step "どう進める？"

• Draw conclusion (supported / rejected / inconclusive)
• Refine hypothesis and re-investigate (→ hypothesize event + /analysis-journal)
• Branch to new hypothesis (→ /analysis-journal branch workflow)

Step 4: Record Conclusion

If user chooses to conclude:

1. Record conclude event in journal:

   - id: "{design_id}-E{nn}"
     type: conclude
     content: "{user's conclusion statement}"
     metadata:
       resolves: ["{question_event_ids that are now resolved}"]
     created_at: "{now}"
   

2. Suggest status transition: "結論が出た。ステータスを変更する？"

   • supported: transition_design_status(design_id, "analyzing", "supported")
   • rejected: transition_design_status(design_id, "analyzing", "rejected")
   • inconclusive: transition_design_status(design_id, "analyzing", "inconclusive")

   Note: Current status must be "analyzing" for terminal transition. If status is "in_review", suggest going through the review process first.

3. Show final summary and suggest:

   • "レビューコメントを残すなら save_review_comment を使う"
   • "この知見をカタログに登録するなら /catalog-register"
   • "派生仮説を立てるなら /analysis-design"

Step 5: Record Refinement (if not concluding)

If user chooses to refine:

1. Record hypothesize event with the refined hypothesis in journal
2. Optionally call update_analysis_design(design_id, hypothesis_statement=refined)
3. Suggest: "/analysis-journal {id} で追加調査を続けよう"

MCP Tool Reference (Existing Only)

Chaining

Language Rules

• Follow project CLAUDE.md language settings. Default to Japanese if no setting.
• Code, IDs, tool names, and YAML fields always stay in English.
• Reflection content follows the user's language (usually Japanese).