| Trigger conditions | Action |
|---|---|
| When a research request is received | Phase 0: Execute project initialization first |
| When the objective is ambiguous | Phase 1: Objective exploration through one-question-one-answer |
| When formulating a plan | Phase 2: Task decomposition and framework recommendations |
⛔ Most important rule
Before starting research, npx shikigami new <ProjectName> must always be executed.
- ❌ Prohibited: Searching the web immediately after receiving a request
- ❌ Prohibited: Creating a folder before confirming the project name
- ✅ Correct: First ask for the project name → wait for the user's approval → Execute CLI
Phase 0: Project initialization
⚠️ Required flow (v1.24.0)
1. プロジェクト名を提案してユーザーに確認
2. ユーザーが「はい」「OK」「承認」等で同意するまで待機
3. 承認後に npx shikigami new <ProjectName> を実行
4. フォルダ作成確認後、ワークフロー開始
❌ Prohibited pattern:
「プロジェクト名はXXXでよろしいでしょうか?」
「確認いただく前に、プロジェクトを作成しておきます」 ← これは禁止!
✅ Correct pattern:
「プロジェクト名はXXXでよろしいでしょうか?」
→ ユーザーの返答を待つ(何も実行しない)
→ ユーザーが承認したら npx shikigami new XXX を実行
Command
npx shikigami new <ProjectName> # 英数字CamelCase、ハイフン禁止
Folders to be created: projects/pjXXXXX_<Name>_YYYYMMDD/
Phase 0-2: Project configuration (v1.25.0) 🆕
After creating the folder,alwaysconfigure the project with the MCP tool:
{"tool":"set_project","arguments":{"autoDetect":true}}
As a result, subsequentsave_prompt/save_researchwill be saved in the project directory.
Phase 0-1: Knowledge inheritance (v1.8.0)
- Search for related reports
- Extract inheritable knowledge (confirmed facts/source URLs)
- User confirmation
Phase 0.5: Structured prompt generation (v1.19.0)
| Element | Description |
|---|---|
| PURPOSE | Objective/Goal |
| TARGET | Research target |
| SCOPE | Scope/Depth |
| TIMELINE | Deadline |
| CONSTRAINTS | Constraints |
| DELIVERABLES | Expected deliverables |
Phase 1: Dialogic objective exploration
⚠️ Strictly one-question-one-answer (collect in 3–7 questions)
| Category | Example questions |
|---|---|
| WHY | What will you decide based on this research? |
| WHO | Who will use the results? |
| WHAT-IF | If the ideal results are obtained, what will you do next? |
| CONSTRAINT | What are the time and budget constraints? |
| SUCCESS | What would make this a success? |
## ❓ 質問 1/5
**カテゴリ**: WHY
この調査で何を判断したいですか?
5 Whys analysis
For abstract answers, dig deeper with "Why?" up to 5 times
JTBD analysis
- Functional job: The task you want to accomplish
- Emotional job: The emotions you want to feel/avoid
- Social job: How others evaluate you
Phase 2: Planning
## 📋 リサーチ計画
| # | タスク | 種別 | 推奨手法 | 優先度 |
|---|--------|------|----------|--------|
| 1 | [タスク] | 調査 | Deep Research | P0 |
| 2 | [タスク] | 分析 | SWOT | P1 |
Success criteria type (v1.6.0)
| Type | Report characteristics |
|---|---|
| IMMEDIATE_APPROVAL | Conclusion first, emphasizing urgency |
| STAGED_APPROVAL | Phased summary |
| CONSIDERATION | Comparison of options |
| BUDGET_APPROVAL | Detailed financial plan, ROI analysis |
| PILOT_APPROVAL | PoC plan, Go/No-Go criteria |
Phase 3: Progress management
Deliverable definition (v1.10.0)
deliverables:
- id: "DEL-001"
type: "report"
status: "in_progress" # not_started/in_progress/completed/blocked
progress: 65
priority: "P0"
Dashboard
[██████████░░░░░░░░░░] 52% 完了
完了:2 / 作業中:2 / ブロック:0 / 未着手:1
Phase 4: Persona information collection (v1.6.0)
| Item | Options |
|---|---|
| Decision-making style | data_driven / intuitive / consensus / balanced |
| Domain knowledge | expert / intermediate / beginner |
| Key focus points | ROI / Risk / Competitive advantage / ESG / Innovation |
→ manifest.yaml personas[] Output to
Automatic invocation (v1.16.0)
| Trigger | Invocation prompt |
|---|---|
| BUDGET_APPROVAL | @risk-sensitivity, @financial-analysis |
| STAGED_APPROVAL | @stakeholder-impact |
| Before plan completion | Proposal checklist validation |
Save destination
projects/pjXXXXX_Name_YYYYMMDD/manifest.yaml
Related skills
- shikigami-deep-research: Research execution based on the plan
- shikigami-consulting-framework: Framework analysis
- shikigami-writing: Final report generation