| name | thought-organizer |
| description | Use when raw, scattered ideas or notes need structure before writing, presenting, or deciding. Triggers on: "생각 정리해줘", "노트 정리", "머릿속이 복잡해", "아이디어 구조화", "outline 만들어줘", mind map", "이거 어떻게 엮어야 해?", "글 쓰기 전에 구조 잡아줘", "산만한 생각 정리". |
| scenarios | ["머릿속에 있는 것들 정리 좀 해줘","이 노트들 구조화해서 아웃라인 만들어줘","I have a bunch of ideas — help me see how they relate","발표 전에 내용 구조 잡아야 해","앱 만들고 싶은데 생각이 너무 산만해","복잡한 개념들 연결해서 정리해줘"] |
| compatibility | {"recommended":["think-tool"],"optional":["sequential-thinking"],"remote_mcp_note":"think-tool이 있으면 구조화 후 모순·누락·미해결 긴장을 체계적으로 검토할 수 있습니다. Claude 설정 → MCP Servers에서 remote SSE 엔드포인트를 추가하세요."} |
| related | ["brainstorming","first-principles"] |
Thought Organizer
Takes messy, scattered input — rough notes, half-formed ideas, stream-of-consciousness text — and produces clear, structured output with hierarchy, connections, and a distilled core message.
Core Workflow
- Absorb — Read all input without filtering. Identify every distinct idea, no matter how fragmented.
- Extract atoms — Pull out the smallest meaningful units (claims, questions, observations, feelings).
- Checkpoint: Have you captured everything, including implicit assumptions? If not, ask one focused clarifying question before continuing.
- Cluster — Group atoms by theme or relationship. Name each cluster with a short label.
- Rank — Identify the single core claim or goal. Mark supporting ideas vs. tangential ones.
- Structure — Choose the right output shape (see Techniques below) and build the hierarchy.
- Surface gaps — Note what is missing, contradictory, or unresolved. Flag explicitly.
- Deliver — Present the structured output, then offer next steps (expand a section, convert to draft, add sources).
MCP Usage
- sequential-thinking: Invoke for steps 3–5 when input contains more than ~5 distinct ideas or the user explicitly requests a structured breakdown. Each step's output should be passed explicitly as input to the next.
- think-tool: Invoke at step 6 (Surface gaps). Use it to enumerate contradictions, missing evidence, and unresolved tensions before delivering output.
Technique Selection
Choose the output shape before structuring. Use this table:
| Situation | Technique |
|---|
| Goal is a document, essay, or presentation | Outline |
| User says "map" or input has no clear hierarchy | Mind Map |
| 5+ ideas with named cross-links or a knowledge-base goal | Zettelkasten |
| User needs only the core message, not full structure | Core Claim Extraction |
Techniques
Outline (default for writing prep)
Use numbered hierarchy. Start with the core claim as the title.
Core Claim / Title
1. First main point
1.1 Supporting detail
1.2 Example
2. Second main point
2.1 ...
3. Counterargument or tension
Conclusion / So what?
Text Mind Map (for visual thinkers)
Use indented branching. Center node first, branches radiate out.
[CENTER: Core topic]
├── Branch A
│ ├── Sub-idea A1
│ └── Sub-idea A2
├── Branch B
│ └── Sub-idea B1
└── Branch C (tension / open question)
Zettelkasten-style Concept Linking
For complex ideas with many cross-connections. Each atom gets an ID and explicit links.
[001] Core claim: ...
[002] Supporting evidence → links to [001]
[003] Counterexample → challenges [001], see also [004]
[004] Resolution or reframe → synthesizes [001] + [003]
[?] Open question: ...
Use this when ideas recur across different contexts or the user wants to build a knowledge base rather than a single document.
Core Claim Extraction
When the user just needs the essence:
핵심 주장: [one sentence]
근거:
- ...
- ...
반론 가능성: ...
다음 질문: ...
Concrete Examples
Example 1 — Messy Input to Outline
Input:
"앱 만들고 싶은데, 사람들이 독서 기록을 남기는 거. 소셜 기능도 있으면 좋겠고, 근데 너무 복잡해질 것 같고, 일단 혼자 쓸 수 있어야 하고, 나중에 친구한테 추천도 하고 싶고, 책 표지 이미지도 넣고 싶어."
Output:
독서 기록 앱
1. 핵심 가치: 개인 독서 이력 관리
1.1 책 검색 및 등록 (표지 이미지 포함)
1.2 독서 상태 추적 (읽는 중 / 완료 / 읽고 싶음)
1.3 메모 및 감상 기록
2. 소셜 기능 (2단계)
2.1 친구에게 책 추천 공유
2.2 독서 피드 (선택적 공개)
3. 미결 사항
- 소셜 기능은 MVP 이후로 분리할 것 (복잡도 관리)
- 책 데이터 소스 결정 필요 (Open Library, 알라딘 API 등)
See references/examples.md for more examples: essay framing and philosophical concept linking.
Output Format Rules
- Always start with the structured output — do not summarize the input back to the user first.
- Label clearly: outline, mind map, concept map, or core claim extraction.
- Flag gaps or contradictions with a dedicated section (미결 / Open Questions).
- Keep language consistent with the user's input language (한국어 입력 → 한국어 출력).
- Offer one concrete next step at the end (e.g., "이 아웃라인으로 초안 써드릴까요?").
Constraints
MUST DO
- Preserve the user's original intent — do not impose a narrative
- Separate facts/claims from feelings/attitudes when both are present
- Make hierarchy explicit (numbering, indentation, IDs)
- Surface contradictions rather than smoothing them over
- Ask at most one clarifying question if input is too sparse to structure
MUST NOT DO
- Do not paraphrase the raw input back without structuring it
- Do not add ideas the user did not express
- Do not pick a structure type without considering what serves the content best
- Do not skip the "gaps" step — unresolved tensions are valuable
Related Skills
brainstorming — 구조화 후 새로운 아이디어나 설계 방향이 필요할 때first-principles — 정리된 아이디어 중 근본 가정을 검증하고 싶을 때
