@nahisaho/shikigami 1.34.0 → 1.37.0

package/AGENTS.md

@@ -1,4 +1,4 @@
-# SHIKIGAMI - Deep Research Agent Skills v1.34.0
+# SHIKIGAMI - Deep Research Agent Skills v1.35.0
 
 > Knowledge for AI Coding Agents — Triggers and Actions defined with WHEN/DO syntax
 
@@ -8,7 +8,7 @@
 
 | Item | Value |
 |------|-------|
-| Version | 1.34.0 |
+| Version | 1.37.0 |
 | Agent Skills | 4 |
 | MCP Tools | 11 |
 | Frameworks | 53 |
@@ -24,9 +24,13 @@
 
 ```
 WHEN: Received a research, investigation, or analysis request
-DO: Always execute `npx shikigami new <ProjectName>` before starting research
+DO: 
+  - If user says "project created" or "created project directory": 
+    → Call set_project(autoDetect: true) and proceed to Phase 1
+  - Otherwise:
+    → Execute `npx shikigami new <ProjectName>` before starting research
 THEN: Call set_project(autoDetect: true) to activate the project
-FORBIDDEN: Web search, using search tools, or gathering information before project creation
+FORBIDDEN: Web search, using search tools, or gathering information before project activation
 ```
 
 ### Prohibition 2: save_research Required After search/visit
@@ -70,20 +74,40 @@ save_prompt(content: <user_input>, type: "original"|"answer"|"instruction"|"feed
 
 ## Workflow Execution Order
 
-When receiving a research request, strictly follow this order:
+**CRITICAL: Follow this order EXACTLY. Do NOT skip steps.**
 
-1. **Decide Project Name** → Confirm with user (CamelCase, alphanumeric only)
-2. **Create Project** → `npx shikigami new <ProjectName>`
+### Case A: User already created project directory
+
+When user says "project created", "created project", or similar:
+
+1. **Set Context** → `set_project(autoDetect: true)`
+2. **Save Original Prompt** → `save_prompt(type: "original")`
+3. **Prompt Optimization** (Phase 1) → Structure into 6 elements → Display → **WAIT for user approval**
+4. **Purpose Discovery** (Phase 2) → Discover true purpose through 1Q1A dialogue
+5. **Deep Research** (Phase 3) → Think→Search→Report→Action cycle
+6. **Framework Analysis** (Phase 4) → Apply SWOT/3C/PEST etc.
+7. **Report Generation** (Phase 5) → Create structured report
+8. **Complete** (Phase 6) → Update manifest.yaml, verify deliverables
+
+### Case B: New research request (no project yet)
+
+When receiving a new research request:
+
+1. **ASK Project Name** → "What project name would you like?" → **WAIT for user response**
+2. **Create Project** → `npx shikigami new <ProjectName>` (only after user confirms)
 3. **Set Context** → `set_project(autoDetect: true)`
 4. **Save Original Prompt** → `save_prompt(type: "original")`
 5. **Verify Folder** → Confirm `projects/pjXXXXX_<Name>_YYYYMMDD/` exists
-6. **Inherit Knowledge** → Search past reports
-7. **Prompt Optimization** (Phase 1) → Structure into 6 elements → User approval
-8. **Purpose Discovery** (Phase 2) → Discover true purpose through 1Q1A dialogue
-9. **Deep Research** (Phase 3) → Think→Search→Report→Action cycle
-10. **Framework Analysis** (Phase 4) → Apply SWOT/3C/PEST etc.
-11. **Report Generation** (Phase 5) → Create structured report
-12. **Complete** (Phase 6) → Update manifest.yaml, verify deliverables
+6. **Prompt Optimization** (Phase 1) → Structure into 6 elements → Display → **WAIT for user approval**
+7. **Purpose Discovery** (Phase 2) → Discover true purpose through 1Q1A dialogue
+8. **Deep Research** (Phase 3) → Think→Search→Report→Action cycle
+9. **Framework Analysis** (Phase 4) → Apply SWOT/3C/PEST etc.
+10. **Report Generation** (Phase 5) → Create structured report
+11. **Complete** (Phase 6) → Update manifest.yaml, verify deliverables
+
+**MANDATORY APPROVAL POINTS** (must wait for user before proceeding):
+- Case B Step 1: Project name confirmation
+- Phase 1 (both cases): Structured prompt approval
 
 ---
 
@@ -94,13 +118,25 @@ When receiving a research request, strictly follow this order:
 ```
 WHEN: Received research request
 DO:
-  1. Confirm project name with user
-  2. Execute npx shikigami new <ProjectName>
-  3. Execute set_project(autoDetect: true)
-  4. Save original prompt with save_prompt(type: "original")
+  Case A - User says "project created" / "created project directory":
+    1. Execute set_project(autoDetect: true)
+    2. Save original prompt with save_prompt(type: "original")
+    3. Proceed to Phase 1 immediately
+  
+  Case B - New request (no project mentioned):
+    1. STOP and ask user: "What project name would you like? (CamelCase, alphanumeric only)"
+    2. WAIT for user response (do NOT proceed without approval)
+    3. After user confirms name: Execute npx shikigami new <ProjectName>
+    4. Execute set_project(autoDetect: true)
+    5. Save original prompt with save_prompt(type: "original")
 OUTPUT: projects/pjXXXXX_ProjectName_YYYYMMDD/
 ```
 
+**CRITICAL**: 
+- If user indicates project is already created → skip to set_project and proceed
+- If new request → MUST ask user for project name confirmation
+- You MUST NOT skip to research phase without project activation
+
 **Naming Rules**:
 - Folder: `pjXXXXX_ProjectName_YYYYMMDD`
 - ProjectName: CamelCase, alphanumeric only (hyphens, spaces, Japanese characters prohibited)
@@ -108,17 +144,25 @@ OUTPUT: projects/pjXXXXX_ProjectName_YYYYMMDD/
 
 ### Phase 1: Prompt Optimization
 
+**THIS PHASE IS MANDATORY - DO NOT SKIP**
+
 ```
 WHEN: Phase 0 completed
 DO:
-  1. Structure original prompt into 6 elements
-  2. Display structured prompt
+  1. Analyze original prompt and structure into 6 elements (see table below)
+  2. Display structured prompt to user in formatted table
   3. Save with save_prompt(type: "structured")
-  4. Wait for user approval
-  5. Save approval/modifications with save_prompt
+  4. Ask user: "Please review and approve this structured prompt, or suggest modifications."
+  5. WAIT for user approval (do NOT proceed without approval)
+  6. Save approval/modifications with save_prompt(type: "approval" or "instruction")
 OUTPUT: 6-element structured prompt (user approved)
 ```
 
+**CRITICAL**:
+- You MUST display the 6-element structured prompt
+- You MUST wait for user approval before proceeding to Phase 2
+- You MUST NOT skip prompt optimization
+
 **6 Elements**:
 | Element | Description |
 |---------|-------------|
@@ -186,22 +230,56 @@ OUTPUT: Framework application results, analysis insights, recommended actions
 | Trigger Keywords | Recommended Frameworks |
 |------------------|------------------------|
 | strategy, strengths, weaknesses, opportunities, threats | SWOT, PEST, 5Forces, 3C |
-| matching, fit, partnership | Matching Matrix |
-| roadmap, plan, phases | Technology Roadmap |
-| customer, user, persona | Persona Analysis, STP |
+| matching, fit, partnership, collaboration | Matching Matrix |
+| roadmap, plan, phases, timeline | Technology Roadmap |
+| customer, user, persona, segment | Persona Analysis, STP, Customer Journey |
+| pricing, price, cost, value | 4P, Value-Based Pricing, Price Sensitivity |
+| market entry, go-to-market, launch | AARRR, GTM Strategy, Ansoff Matrix |
+| competitive, differentiation, positioning | Blue Ocean, Positioning Map, Competitive Analysis |
+| process, efficiency, workflow, optimization | Value Chain, ECRS, Business Process |
+| decision, prioritization, evaluation | Decision Matrix, Eisenhower, Weighted Scoring |
+| business model, revenue, monetization | Business Model Canvas, Lean Canvas, Revenue Model |
+| root cause, problem, issue, why | Fishbone (Ishikawa), 5 Whys, Issue Tree |
+| organization, team, structure, roles | McKinsey 7S, RACI, Org Design |
+| risk, uncertainty, mitigation | Risk Matrix, FMEA, Scenario Planning |
+| innovation, idea, creativity | Design Thinking, SCAMPER, Brainstorming |
+| growth, scale, expansion | BCG Matrix, GE Matrix, Product Lifecycle |
+
+**CRITICAL**: Select framework based on research PURPOSE, not just keywords. When multiple frameworks apply, prioritize:
+1. Frameworks that directly answer the user's core question
+2. Frameworks with available data from Phase 3 research
+3. Simpler frameworks over complex ones when depth is similar
 
 ### Phase 5: Report Generation
 
 ```
 WHEN: Phase 4 completed, 1+ framework applied
 DO:
-  1. Determine report structure (research report/proposal/Qiita article etc.)
-  2. Write content (with hallucination prevention check)
-  3. Organize citations and references (URL required for all data)
-  4. Quality verification (numerical consistency, executive summary sync)
+  1. Select appropriate report template (see Template Selection below)
+  2. Write content following template structure
+  3. Hallucination prevention check (verify all facts against sources)
+  4. Organize citations and references (URL required for all data)
+  5. Quality verification (numerical consistency, executive summary sync)
 OUTPUT: Completed report (under reports/), reference list
 ```
 
+**Template Selection Criteria**:
+| Report Type | Template | Use When |
+|-------------|----------|----------|
+| Research Report | research-report | General investigation, market research, technology assessment |
+| Proposal | proposal | Recommendations, action plans, business cases |
+| Qiita Article | qiita-article | Technical content for publication |
+| Executive Summary | executive-summary | Brief overview for decision makers |
+| Comparison Report | comparison-report | Product/service/technology comparisons |
+| Feasibility Study | feasibility-study | Viability assessment, go/no-go decisions |
+| Competitive Analysis | competitive-analysis | Competitor landscape, positioning |
+| Technical Report | technical-report | Deep technical investigation |
+
+**CRITICAL**: Always use `shikigami-report-template` skill to load the appropriate template BEFORE writing. Template provides:
+- Required sections and structure
+- Formatting guidelines
+- Quality checklist
+
 ---
 
 ## MCP Tools List
@@ -414,4 +492,4 @@ DO: Auto-detect Shift_JIS/EUC-JP/ISO-2022-JP and convert to UTF-8
 ---
 
 **Last Updated**: 2026-01-28  
-**Version**: 1.33.0
+**Version**: 1.37.0

package/CHANGELOG.md

@@ -5,6 +5,39 @@ All notable changes to SHIKIGAMI will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.37.0] - 2026-01-28
+
+### Added
+
+- **AGENTS.md: Enhanced framework selection criteria**
+  - Expanded from 4 to 15 framework categories
+  - Added: pricing, market entry, competitive, process, decision, business model, root cause, organization, risk, innovation, growth
+  - Added CRITICAL note for framework selection prioritization
+
+- **AGENTS.md: Report template selection guidance**
+  - Added Template Selection Criteria table (8 report types)
+  - Added instruction to use `shikigami-report-template` skill before writing
+  - Updated Phase 5 DO steps to include template selection
+
+### Changed
+
+- **AGENTS.md: Case A workflow for pre-created projects** (from v1.36.0)
+  - Added recognition of "project created" statements
+  - Split workflow into Case A (project exists) and Case B (new request)
+  - Phase 0 now handles both scenarios
+
+## [1.35.0] - 2026-01-28
+
+### Changed
+
+- **AGENTS.md: Enforce mandatory approval points**
+  - Phase 0: Added explicit "ASK user" and "WAIT for response" instructions
+  - Phase 0: Added CRITICAL notes preventing skip without user approval
+  - Phase 1: Added "THIS PHASE IS MANDATORY - DO NOT SKIP" header
+  - Phase 1: Added explicit "WAIT for user approval" instruction
+  - Workflow: Added "MANDATORY APPROVAL POINTS" section listing required user confirmations
+  - Workflow: Changed "Decide Project Name" to "ASK Project Name" with explicit wait
+
 ## [1.34.0] - 2026-01-28
 
 ### Changed

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nahisaho/shikigami",
-  "version": "1.34.0",
+  "version": "1.37.0",
   "description": "GitHub Copilot Agent Skills for Deep Research & Consulting - AI-Powered Research Assistant with 50+ Consulting Frameworks",
   "keywords": [
     "github-copilot",