agileflow 3.4.3 → 4.0.0-alpha.2

package/src/core/commands/research/synthesize.md

@@ -1,928 +0,0 @@
----
-description: Synthesize insights across multiple research files
-argument-hint: "[TOPIC=<text>] [FILES=<list>]"
-compact_context:
-  priority: high
-  preserve_rules:
-    - "ACTIVE COMMAND: /agileflow:research:synthesize - Cross-research synthesis"
-    - "Search docs/10-research/ for matching files by keyword or use provided FILES list"
-    - "Extract metadata: date, findings, approach, sources, related items"
-    - "Find CONSENSUS (2+ files agree = HIGH), UNIQUE (1 file = verify), CONFLICTS (disagreement = flag)"
-    - "Show synthesis BEFORE recommending artifacts (diff-first)"
-    - "Use confidence scoring: HIGH (2+ agree), UNIQUE (1 file), CONFLICT (disagree)"
-    - "After synthesis, offer Implementation Ideation with multi-expert analysis (like /research:import)"
-    - "If ideation requested: EnterPlanMode → deploy 3-5 experts IN PARALLEL → synthesize → Implementation Ideation Report"
-    - "Intelligent artifact recommendation based on research type (not always ADR/Epic)"
-    - "Always end with AskUserQuestion for next steps"
-  state_fields:
-    - selected_files
-    - metadata_extracted
-    - consensus_findings
-    - conflicts_found
-    - analysis_requested
-    - experts_deployed
-    - expert_results
-    - consensus_level
----
-
-# /agileflow:research:synthesize
-
-Synthesize insights across multiple research files to find patterns, detect conflicts, and aggregate knowledge.
-
----
-
-## Purpose
-
-Query across your research notes to discover:
-- **Consensus findings** - What do multiple research files agree on? (HIGH confidence)
-- **Unique insights** - What appears in only one file? (Specialized or outdated?)
-- **Conflicts** - Where do research files disagree? (Needs human review)
-- **Patterns** - Common technologies, timelines, categories across research
-
-**RLM Concept Applied**: This implements the "dependency graph over documents" pattern - treating research files as nodes in a graph and finding cross-document relationships.
-
----
-
-## STEP 0: Gather Context
-
-```bash
-node .agileflow/scripts/obtain-context.js research:synthesize
-```
-
----
-
-<!-- COMPACT_SUMMARY_START -->
-
-## ⚠️ COMPACT SUMMARY - /agileflow:research:synthesize IS ACTIVE
-
-**CRITICAL**: You are running `/agileflow:research:synthesize`. This command queries across multiple research files and optionally analyzes for implementation.
-
-**ROLE**: Find patterns, consensus, and conflicts across research. Apply confidence scoring. Optionally deploy multi-expert analysis for implementation ideation.
-
----
-
-### 🚨 RULE #1: VALIDATE INPUTS FIRST
-
-**Require TOPIC or FILES:**
-- `TOPIC=<text>` - Search research files by keyword
-- `FILES=<comma-separated>` - Specific files to analyze
-
-If neither provided, ask user:
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "What topic would you like to synthesize research about?",
-  "header": "Topic",
-  "multiSelect": false,
-  "options": [
-    {"label": "Enter topic keyword", "description": "Search research files by keyword (use 'Other' to type)"},
-    {"label": "List all research first", "description": "Show research index to pick files"},
-    {"label": "Analyze ALL research", "description": "Synthesize entire research library (may be slow)"}
-  ]
-}]</parameter>
-</invoke>
-```
-
----
-
-### 🚨 RULE #2: SEARCH OR SELECT RESEARCH FILES
-
-**If TOPIC provided:**
-1. Search `docs/10-research/*.md` AND `docs/10-research/prompts/*.md` for files containing keyword
-2. Check filename, title (# header), key findings, summary
-3. Present matching files to user for confirmation
-
-**If FILES provided:**
-1. Validate each file exists in `docs/10-research/` or `docs/10-research/prompts/`
-2. Warn if any file not found in either location
-3. Proceed with valid files
-
-**Auto-select if <5 matches. Ask user to confirm if 5+ matches.**
-
----
-
-### 🚨 RULE #3: EXTRACT METADATA FROM EACH FILE
-
-For each selected research file, extract:
-
-| Field | Source |
-|-------|--------|
-| Date | Filename (YYYYMMDD) or **Date**: header |
-| Topic | First `# ` header |
-| Status | **Status**: field (Active/Superseded/Archived) |
-| Summary | ## Summary section (first paragraph) |
-| Key Findings | Bullet list under ## Key Findings |
-| Recommended Approach | ## Recommended Approach section |
-| Technologies | Keywords in findings (React, Next.js, etc.) |
-| Related ADRs | ## Related section ADR links |
-| Related Stories | ## Related section story links |
-| Sources | ## Sources section URLs |
-| Age Days | Today - file date |
-
----
-
-### 🚨 RULE #4: APPLY CONFIDENCE SCORING
-
-| Category | Definition | Threshold | Action |
-|----------|------------|-----------|--------|
-| **HIGH** | Finding appears in 2+ files | 2+ matches | Strong recommendation |
-| **MEDIUM** | Related findings, not exact | Similar topics | Consider context |
-| **UNIQUE** | Only in 1 file | 1 match | Verify: newer insight or outdated? |
-| **CONFLICT** | Files disagree on approach | Contradictory | Flag for human review |
-
-**How to detect:**
-- **Consensus**: Same finding/approach mentioned in multiple files
-- **Unique**: Grep for finding - only 1 file contains it
-- **Conflict**: Different recommendations for same topic (e.g., "use Redis" vs "use in-memory cache")
-
----
-
-### 🚨 RULE #5: SHOW SYNTHESIS BEFORE ARTIFACTS (DIFF-FIRST)
-
-**Present synthesis report BEFORE offering to create anything:**
-
-```markdown
-## Research Synthesis: [TOPIC]
-
-### Files Analyzed (N total)
-| Date | Title | Status |
-|------|-------|--------|
-| YYYY-MM-DD | [Topic] | Active |
-
-### Consensus Findings (HIGH CONFIDENCE)
-Findings that appear in 2+ research notes:
-- **[Finding]** - Sources: [file1.md, file2.md]
-  - [Supporting detail from files]
-
-### Unique Insights (VERIFY)
-Findings only in one research note:
-- **[Finding]** - Source: [file.md]
-  - Age: N days | Reason: [Specialized topic / Newer research / Consider updating]
-
-### Conflicts Detected (NEEDS REVIEW)
-Different approaches recommended:
-- **[Topic]**:
-  - [file1.md]: Recommends X because [reason]
-  - [file2.md]: Recommends Y because [reason]
-  - **Resolution needed**: [suggestion]
-
-### Technology Patterns
-| Technology | Files | First Mentioned | Most Recent |
-|------------|-------|-----------------|-------------|
-| Next.js | 5 | 2025-12-01 | 2026-01-17 |
-
-### Related Artifacts
-- **ADRs**: [list of ADRs mentioned across files]
-- **Stories**: [list of stories mentioned]
-- **Epics**: [list of epics mentioned]
-
-### Timeline
-[Oldest research date] → [Newest research date]
-```
-
----
-
-### 🚨 RULE #6: END WITH AskUserQuestion (OFFER IDEATION)
-
-**After showing synthesis, offer implementation ideation (like /research:import):**
-
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "Synthesis of '[TOPIC]': [N] consensus findings, [M] conflicts across [K] files. What next?",
-  "header": "Next step",
-  "multiSelect": false,
-  "options": [
-    {"label": "Implementation ideation with [detected-type] experts (Recommended)", "description": "Deploy [3-5] experts ([expert list]) to analyze [N] consensus findings for your codebase"},
-    {"label": "Save synthesis report", "description": "Save to docs/10-research/[YYYYMMDD]-synthesis-[slug].md"},
-    {"label": "Flag [M] conflicts for review", "description": "Mark conflicting research on [conflict topics] for update"},
-    {"label": "Narrow scope - different topic", "description": "Current: '[TOPIC]' matched [K] files - try different keywords"},
-    {"label": "Done", "description": "Synthesis complete, [N] consensus + [M] conflicts documented"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-**Key**: Populate `[TOPIC]` with search topic, `[N]` with consensus finding count, `[M]` with conflict count, `[K]` with files analyzed count. For ideation, auto-detect expert type from synthesis content and list specific experts (e.g., "security, api, testing"). For "Flag conflicts", reference the actual conflict topics found. If no conflicts found (M=0), replace that option with "Synthesize with additional files" or remove it.
-
----
-
-### 🚨 RULE #7: IMPLEMENTATION IDEATION = PLAN MODE + MULTI-EXPERT ANALYSIS
-
-**When user requests ideation, you MUST:**
-
-1. **Enter plan mode**:
-```xml
-<invoke name="EnterPlanMode"/>
-```
-
-2. **Gather project context**:
-```bash
-node .agileflow/scripts/obtain-context.js babysit
-```
-
-3. **Deploy 3-5 domain experts IN PARALLEL** (see RULE #9 for expert selection)
-
-4. **Synthesize results** with confidence scoring
-
-**Why**: Single-perspective analysis misses domain-specific considerations. Multi-expert analysis provides security, performance, testing, and architecture perspectives for comprehensive implementation guidance.
-
----
-
-### 🚨 RULE #8: SHOW BENEFITS FIRST IN ANALYSIS
-
-**Order matters: Benefits → Implementation → Complexity → Risks.**
-
-```
-❌ WRONG: "We'd modify 5 files, add 2 dependencies, refactor the system..."
-✅ RIGHT: "You'd gain: 40% faster auth, better security, cleaner codebase.
-           To implement, we'd modify 5 files, add 2 dependencies..."
-```
-
----
-
-### 🚨 RULE #9: IMPLEMENTATION IDEATION (Multi-Expert Analysis)
-
-When user requests "Implementation ideation with multi-expert analysis":
-
-1. **Enter Plan Mode** and auto-detect research type to select experts
-2. **Deploy 3-5 experts IN PARALLEL** (like /ideate pattern)
-3. **Each expert analyzes from their domain perspective**:
-   - Implementation fit for codebase
-   - Domain-specific considerations
-   - Recommended approach with specific files
-   - Risks and gotchas
-4. **Synthesize with confidence scoring**:
-   - HIGH CONFIDENCE: 2+ experts agree on approach/files
-   - MEDIUM CONFIDENCE: 1 expert with specific evidence
-5. **Present unified Implementation Ideation Report** with expert consensus
-
-**Expert Selection by Research Type:**
-
-| Research Type | Experts (Deploy ALL in parallel) |
-|--------------|--------------------------------|
-| Architecture/Framework | api, database, performance, security (4) |
-| Feature Implementation | api, ui, testing, database (4) |
-| Security/Compliance | security, api, testing, compliance (4) |
-| Performance | performance, database, api, monitoring (4) |
-| Best Practices | refactor, testing, documentation (3) |
-| Full-Stack (default) | api, ui, database, testing, security (5) |
-
-**Detection Keywords:**
-- **Security**: auth, oauth, jwt, encryption, vulnerability, compliance
-- **Performance**: cache, optimize, latency, throughput, benchmark
-- **Architecture**: migrate, upgrade, framework, refactor, redesign
-- **UI**: component, styling, accessibility, ux, design system
-- **Database**: schema, migration, query, index, model
-
----
-
-### 🚨 RULE #10: INTELLIGENT ARTIFACT SELECTION
-
-**Recommend artifact type based on expert consensus. DON'T default to ADR or Epic.**
-
-| Research Type | Artifact | Indicators |
-|---|---|---|
-| Architecture/tech decision | **ADR** | Trade-offs, "use X or Y?", one-time decision |
-| Large feature (5+ steps) | **Epic + Stories** | Multiple files, multiple domains, 3+ days |
-| Single focused task | **Story** | 1-3 files, 1-4 hours, single domain |
-| Best practices/guidelines | **Practice doc** | "How to do X", guidelines, no feature work |
-| Code quality/refactoring | **Tech debt item** | No user-facing change, improvement |
-
-**Example recommendations:**
-- "Upgrade to Next.js 15" → ADR
-- "Add OAuth" → Epic + Stories (multiple domains)
-- "Fix cache bug" → Story (single issue)
-
----
-
-### ANTI-PATTERNS (DON'T DO THESE)
-
-❌ Skip validation of TOPIC/FILES
-❌ Include files that don't match the topic
-❌ Show generic synthesis without reading actual file contents
-❌ Jump to artifact creation without showing synthesis first
-❌ Ignore conflicts - they must be flagged
-❌ Treat all findings as equal - use confidence scoring
-❌ Single-perspective analysis (shallow, misses domain considerations)
-❌ Deploy experts one at a time (MUST be parallel)
-❌ Assume ADR is the right artifact for all synthesis
-❌ End without AskUserQuestion
-
-### DO THESE INSTEAD
-
-✅ Validate TOPIC or FILES first
-✅ Search and confirm file selection
-✅ Read and extract metadata from each file
-✅ Apply confidence scoring (HIGH/UNIQUE/CONFLICT)
-✅ Show full synthesis before offering artifacts
-✅ Flag conflicts for human review
-✅ Offer implementation ideation with multi-expert analysis
-✅ Deploy 3-5 domain experts IN PARALLEL for multi-perspective analysis
-✅ Synthesize expert results with confidence scoring
-✅ Recommend artifact type based on expert consensus (not always ADR)
-✅ Always end with AskUserQuestion
-
----
-
-### WORKFLOW
-
-**Phase 1: Input Validation**
-1. Check if TOPIC or FILES provided
-2. If neither, ask user via AskUserQuestion
-3. Store selected topic/files in state
-
-**Phase 2: File Selection**
-4. If TOPIC: Search docs/10-research/*.md for keyword matches
-5. If FILES: Validate each file exists
-6. Present matches, confirm selection if 5+ files
-
-**Phase 3: Metadata Extraction**
-7. For each file, read and extract:
-   - Date, topic, status, summary
-   - Key findings, recommended approach
-   - Technologies, related items, sources
-
-**Phase 4: Synthesis Analysis**
-8. Compare findings across files
-9. Identify consensus (2+ files agree)
-10. Identify unique insights (1 file only)
-11. Detect conflicts (disagreements)
-12. Extract technology patterns
-
-**Phase 5: Report Generation**
-13. Format synthesis report with sections:
-    - Files Analyzed
-    - Consensus Findings (HIGH)
-    - Unique Insights (VERIFY)
-    - Conflicts (NEEDS REVIEW)
-    - Technology Patterns
-    - Related Artifacts
-    - Timeline
-
-**Phase 6: Present & Offer Ideation**
-14. Show synthesis report to user
-15. Ask what to do next via AskUserQuestion (offer ideation)
-
-**Phase 7: Implementation Ideation (If Requested)**
-16. Enter plan mode + select 3-5 domain experts
-17. Deploy experts IN PARALLEL (Task tool with run_in_background)
-18. Collect results (TaskOutput) + synthesize with confidence scoring
-19. Present Implementation Ideation Report
-20. Confirm interest in proceeding
-21. Recommend artifact type based on expert consensus
-22. Create artifact if user confirms
-
----
-
-### KEY FILES
-
-| File | Purpose |
-|------|---------|
-| `docs/10-research/` | Directory containing all research notes |
-| `docs/10-research/README.md` | Index of research with metadata |
-| `docs/10-research/*.md` | Individual research files to analyze |
-
----
-
-### REMEMBER AFTER COMPACTION
-
-- `/agileflow:research:synthesize` IS ACTIVE - cross-research analysis
-- Validate TOPIC or FILES first
-- Search and select relevant research files
-- Extract metadata from each file
-- Apply confidence scoring: HIGH (2+ agree), UNIQUE (1 file), CONFLICT (disagree)
-- Show synthesis BEFORE offering artifacts
-- Flag conflicts for human review
-- Offer Implementation Ideation with multi-expert analysis (like /research:import)
-- If ideation requested: enter plan mode + deploy 3-5 domain experts IN PARALLEL
-- Synthesize expert results with confidence scoring (HIGH = 2+ agree)
-- Present Implementation Ideation Report with expert consensus
-- Recommend artifact type based on expert consensus (not always ADR)
-- End with AskUserQuestion for next steps
-
-<!-- COMPACT_SUMMARY_END -->
-
----
-
-## Arguments
-
-| Argument | Required | Description |
-|----------|----------|-------------|
-| TOPIC | No* | Keyword to search research files (*one of TOPIC or FILES required) |
-| FILES | No* | Comma-separated list of specific files to analyze |
-
-**Examples:**
-```
-/agileflow:research:synthesize TOPIC="authentication"
-/agileflow:research:synthesize TOPIC="performance"
-/agileflow:research:synthesize FILES="20260117-rlm.md,20260109-rag.md"
-```
-
----
-
-## IMMEDIATE ACTIONS
-
-Upon invocation, execute these steps:
-
-### Step 1: Create Todo List
-
-```xml
-<invoke name="TaskCreate/TaskUpdate">
-<parameter name="todos">[
-  {"content": "Validate TOPIC or FILES input", "status": "in_progress", "activeForm": "Validating inputs"},
-  {"content": "Search/select research files", "status": "pending", "activeForm": "Selecting files"},
-  {"content": "Extract metadata from each file", "status": "pending", "activeForm": "Extracting metadata"},
-  {"content": "Analyze for consensus/conflicts", "status": "pending", "activeForm": "Analyzing patterns"},
-  {"content": "Generate synthesis report", "status": "pending", "activeForm": "Generating report"},
-  {"content": "Present synthesis and offer next steps", "status": "pending", "activeForm": "Presenting synthesis"}
-]</parameter>
-</invoke>
-```
-
-### Step 2: Validate Inputs
-
-**If TOPIC provided:**
-- Store topic keyword
-- Proceed to search
-
-**If FILES provided:**
-- Parse comma-separated list
-- Validate each file exists in `docs/10-research/`
-- Proceed with valid files
-
-**If NEITHER provided:**
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "What topic would you like to synthesize research about?",
-  "header": "Topic",
-  "multiSelect": false,
-  "options": [
-    {"label": "Enter topic keyword", "description": "Search research files by keyword (use 'Other' to type)"},
-    {"label": "List all research first", "description": "Show research index to pick files"},
-    {"label": "Analyze ALL research", "description": "Synthesize entire research library"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-### Step 3: Search/Select Research Files
-
-**If TOPIC search:**
-1. List all files in `docs/10-research/` and `docs/10-research/prompts/` (if exists)
-2. For each file, check if topic keyword appears in:
-   - Filename
-   - First `# ` header (title)
-   - `## Summary` section
-   - `## Key Findings` section
-3. Collect matching files
-
-**Present matches:**
-```
-Found N research files matching "[TOPIC]":
-
-| # | Date | Title | Summary |
-|---|------|-------|---------|
-| 1 | 2026-01-17 | RLM - Recursive Language Models | Breakthrough approach... |
-| 2 | 2026-01-09 | Training vs RAG | Three approaches... |
-```
-
-**If 5+ matches, ask for confirmation:**
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "Found N matching files. Analyze all or select specific ones?",
-  "header": "Selection",
-  "multiSelect": false,
-  "options": [
-    {"label": "Analyze all N files (Recommended)", "description": "Comprehensive synthesis"},
-    {"label": "Let me pick specific files", "description": "I'll tell you which ones"},
-    {"label": "Narrow search - different keyword", "description": "Try more specific topic"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-### Step 4: Extract Metadata
-
-For each selected file, read and extract:
-
-```javascript
-// Pseudocode for metadata extraction
-{
-  path: "20260117-rlm.md",
-  date: "2026-01-17",  // From filename YYYYMMDD
-  title: "RLM - Recursive Language Models",  // From # header
-  status: "Active",  // From **Status**: field
-  summary: "Breakthrough approach for AI agents...",  // First paragraph of ## Summary
-  key_findings: [
-    "Context rot is function of context AND task complexity",
-    "REPL + recursion enables intelligent search",
-    "Model documents as dependency graphs"
-  ],
-  recommended_approach: "Use REPL with recursion for complex documents",
-  technologies: ["AI", "LLM", "Python", "REPL"],
-  related_adrs: ["ADR-0008"],
-  related_stories: [],
-  sources: ["Video transcript"],
-  age_days: 0
-}
-```
-
-### Step 5: Synthesis Analysis
-
-**5a. Find Consensus (HIGH confidence)**
-- Group findings by similarity
-- Mark as HIGH if finding appears in 2+ files
-- Example: "Use caching" in file1 + "Add cache layer" in file2 = consensus
-
-**5b. Identify Unique Insights**
-- Findings only appearing in 1 file
-- Check age: older = potentially outdated, newer = potentially valuable new insight
-- Flag for verification
-
-**5c. Detect Conflicts**
-- Different recommendations for same topic
-- Example: "Use Redis" vs "Use in-memory cache" for caching
-- Flag with both positions and reasons
-
-**5d. Extract Patterns**
-- Common technologies across files
-- Timeline of research (oldest → newest)
-- Related artifacts mentioned
-
-### Step 6: Generate Synthesis Report
-
-Format the synthesis:
-
-```markdown
-## Research Synthesis: [TOPIC]
-
-**Generated**: YYYY-MM-DD
-**Files Analyzed**: N research notes
-**Date Range**: [oldest] → [newest]
-
----
-
-### Files Analyzed
-
-| Date | Title | Status | Age |
-|------|-------|--------|-----|
-| 2026-01-17 | RLM - Recursive Language Models | Active | 0 days |
-| 2026-01-09 | Training vs RAG | Active | 8 days |
-
----
-
-### Consensus Findings (HIGH CONFIDENCE)
-
-These findings appear in 2+ research notes:
-
-1. **[Finding Statement]**
-   - Sources: file1.md (line X), file2.md (line Y)
-   - Context: [Brief explanation of agreement]
-
-2. **[Finding Statement]**
-   - Sources: file1.md, file3.md
-   - Context: [Brief explanation]
-
----
-
-### Unique Insights (VERIFY)
-
-Findings that appear in only one research note:
-
-1. **[Finding Statement]** - Source: file.md
-   - Age: N days
-   - Assessment: [Specialized topic / Newer research / Consider reviewing]
-
----
-
-### Conflicts Detected (NEEDS REVIEW)
-
-Different approaches recommended for the same topic:
-
-1. **[Topic: Caching Strategy]**
-   | File | Recommendation | Reason |
-   |------|----------------|--------|
-   | file1.md | Use Redis | Scalability for distributed systems |
-   | file2.md | Use in-memory | Simplicity for single-node |
-
-   **Resolution suggestion**: Consider your deployment architecture
-
----
-
-### Technology Patterns
-
-| Technology | Mentions | First Seen | Most Recent |
-|------------|----------|------------|-------------|
-| Next.js | 5 files | 2025-12-01 | 2026-01-17 |
-| React | 4 files | 2025-12-01 | 2026-01-15 |
-
----
-
-### Related Artifacts
-
-**ADRs mentioned**: ADR-0001, ADR-0003, ADR-0008
-**Stories mentioned**: US-0042, US-0055
-**Epics mentioned**: EP-0004, EP-0007
-
----
-
-### Timeline
-
-```
-2025-12-01  ────────────────────────────  2026-01-17
-   │                                          │
-   ├─ First research: [topic]                 │
-   │                                          │
-   └──────────────────────────────────────────┴─ Latest: [topic]
-```
-```
-
-### Step 7: Present and Offer Next Steps
-
-Show the full synthesis report, then:
-
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "Synthesis of '[TOPIC]': [N] consensus findings, [M] conflicts across [K] files. What next?",
-  "header": "Next step",
-  "multiSelect": false,
-  "options": [
-    {"label": "Implementation ideation with [detected-type] experts (Recommended)", "description": "Deploy [3-5] experts ([expert list]) to analyze [N] consensus findings for your codebase"},
-    {"label": "Save synthesis report", "description": "Save to docs/10-research/[YYYYMMDD]-synthesis-[slug].md"},
-    {"label": "Flag [M] conflicts for review", "description": "Mark conflicting research on [conflict topics] for update"},
-    {"label": "Try different topic", "description": "Current: '[TOPIC]' matched [K] files - try different keywords"},
-    {"label": "Done", "description": "Synthesis complete, [N] consensus + [M] conflicts documented"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-**Key**: Same smart population rules as RULE #6 above. Populate all `[bracketed]` values with actual data from the synthesis. If no conflicts (M=0), replace "Flag conflicts" with "Synthesize with additional files". Auto-detect expert type from synthesis findings.
-
-**If "Implementation ideation"**: Continue to Step 8.
-**If "Save synthesis report"**: Save and update index (see Saving section below).
-**If other**: Handle accordingly.
-
----
-
-### Step 8: Enter Plan Mode and Select Experts
-
-```xml
-<invoke name="EnterPlanMode"/>
-```
-
-Then immediately gather project context:
-
-```bash
-node .agileflow/scripts/obtain-context.js babysit
-```
-
-**Auto-detect research type and select experts:**
-
-Analyze the TOPIC and synthesis findings to determine research type, then select 3-5 experts:
-
-| Research Type | Detection Keywords | Experts |
-|--------------|-------------------|---------|
-| **Security/Compliance** | auth, oauth, jwt, encryption, vulnerability, compliance, csrf, xss | security, api, testing, compliance |
-| **Performance** | cache, optimize, latency, throughput, benchmark, profiling | performance, database, api, monitoring |
-| **Architecture/Framework** | migrate, upgrade, framework, refactor, redesign, architecture | api, database, performance, security |
-| **UI/Frontend** | component, styling, accessibility, ux, design system, responsive | ui, api, testing, accessibility |
-| **Database** | schema, migration, query, index, model, postgres, mongo | database, api, performance, datamigration |
-| **Full-Stack (default)** | multiple domains or unclear | api, ui, database, testing, security |
-
----
-
-### Step 9: Deploy Experts in Parallel
-
-**CRITICAL**: Deploy ALL selected experts in a SINGLE message using the Task tool.
-
-Each expert receives this prompt template:
-
-```
-EXPERTISE FIRST: Read your expertise.yaml file if it exists at .agileflow/expertise/{domain}.yaml
-
-RESEARCH SYNTHESIS: {synthesis report summary}
-CONSENSUS FINDINGS: {list of HIGH confidence findings}
-CONFLICTS: {list of conflicts found}
-PROJECT CONTEXT: {from obtain-context.js output}
-
-TASK: Analyze this research synthesis from your {DOMAIN} perspective:
-
-1. **Implementation Fit**: How well do the consensus findings fit the current codebase?
-   - What patterns/files would need to change?
-   - Any conflicts with existing architecture?
-
-2. **Domain-Specific Considerations**: From your expertise:
-   - For SECURITY: What security implications? Risks? Compliance needs?
-   - For PERFORMANCE: What performance impacts? Bottlenecks? Optimization opportunities?
-   - For TESTING: What test coverage needed? Edge cases? Regression risks?
-   - For API: What API changes? Breaking changes? Versioning needs?
-   - For UI: What UX considerations? Accessibility? Design patterns?
-   - For DATABASE: What schema changes? Migration strategy? Query impact?
-
-3. **Implementation Approach**: Your recommended approach:
-   - Key files to modify (specific paths from codebase)
-   - Effort estimate (hours/days/weeks)
-   - Dependencies or prerequisites
-
-4. **Risks & Gotchas**: What could go wrong?
-   - Technical risks
-   - Migration complexity
-   - Team adoption concerns
-
-FORMAT as structured markdown with specific file paths and evidence from codebase.
-```
-
----
-
-### Step 10: Collect Results and Synthesize
-
-Collect all expert outputs using TaskOutput, then synthesize:
-
-1. **Find Consensus**: Group similar recommendations across experts
-   - **HIGH CONFIDENCE**: 2+ experts recommend same approach/files
-   - **MEDIUM CONFIDENCE**: 1 expert with specific evidence
-
-2. **Aggregate Effort Estimates**: Average across experts, note range
-
-3. **Collect All Risks**: Union of risks from all experts, prioritize by frequency
-
-4. **Validate Artifact Recommendation** via expert consensus:
-   - If experts mention "decision", "tradeoffs", "alternatives" → **ADR**
-   - If experts identify 5+ files across multiple domains → **Epic + Stories**
-   - If experts agree on single focus area, 1-3 files → **Story**
-   - If experts focus on guidelines, patterns → **Practice doc**
-
----
-
-### Step 11: Present Implementation Ideation Report
-
-```markdown
-## Implementation Ideation Report
-
-**Research Synthesis**: {TOPIC}
-**Experts Consulted**: {list of 3-5 experts with badges}
-**Consensus Level**: {High/Medium/Low based on agreement}
-
----
-
-### High-Confidence Implementation Steps
-*Agreed by 2+ experts*
-
-1. **{Step title}**
-   - Experts: {badges}
-   - Files: `path/to/file1.ts`, `path/to/file2.ts`
-   - Effort: {averaged estimate}
-   - Approach: {consensus approach}
-
----
-
-### Domain-Specific Considerations
-
-**Security** (security expert):
-- {key consideration}
-- Mitigation: {recommended approach}
-
-**Performance** (performance expert):
-- {optimization opportunity}
-
-**Testing** (testing expert):
-- {coverage requirements}
-- {edge cases}
-
----
-
-### Risks & Gotchas
-
-| Risk | Expert(s) | Severity | Mitigation |
-|------|-----------|----------|------------|
-| {risk 1} | security | High | {mitigation} |
-
----
-
-### Effort Estimate Summary
-
-| Expert | Estimate | Notes |
-|--------|----------|-------|
-| {expert} | {estimate} | {notes} |
-| **Consensus** | **{averaged}** | {range} |
-
----
-
-### Recommended Artifact
-
-Based on expert consensus: **{ADR/Epic/Story/Practice}**
-**Reason**: {why this artifact type}
-```
-
----
-
-### Step 12: Confirm and Create Artifact
-
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "[N] experts analyzed '[TOPIC]' ([CONSENSUS_LEVEL] consensus). Proceed with implementation?",
-  "header": "Proceed?",
-  "multiSelect": false,
-  "options": [
-    {"label": "Yes - Create [RECOMMENDED_ARTIFACT] (Recommended)", "description": "[N] high-confidence steps identified across [M] files"},
-    {"label": "Modify approach first", "description": "Adjust [specific area of disagreement or concern]"},
-    {"label": "Save analysis to synthesis file", "description": "Append ideation report to [FILENAME] for later"},
-    {"label": "Cancel", "description": "Exit plan mode, synthesis already shown"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-**Key**: Populate `[N]` with expert count, `[TOPIC]` with research topic, `[CONSENSUS_LEVEL]` with High/Medium/Low. For "Create", use the actual recommended artifact type. For "Modify", reference specific disagreement areas from expert results.
-
-**If "Yes"**: Present intelligent artifact selection:
-
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "Expert consensus recommends [ARTIFACT_TYPE] for '[TOPIC]'. What would you like to create?",
-  "header": "Create",
-  "multiSelect": false,
-  "options": [
-    {"label": "[Recommended artifact]: [suggested title] (Recommended)", "description": "[N] experts agreed - [brief reason from consensus]"},
-    {"label": "Create Epic + Stories instead", "description": "Break into [estimated N] stories across [domains detected]"},
-    {"label": "Create single Story instead", "description": "Track as one work item (~[effort estimate])"},
-    {"label": "Create Practice doc instead", "description": "Document as guidelines in docs/02-practices/"},
-    {"label": "Skip artifact creation", "description": "Synthesis + ideation report saved, implement later"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-**Key**: First option MUST be the actual expert-recommended artifact with a suggested title. Include expert count and brief consensus reason. For alternatives, populate with specifics from expert analysis. Remove the recommended artifact from the alternatives list.
-
-**If "Save analysis only"**: Append Implementation Ideation Report to the synthesis file, exit plan mode.
-**If "Cancel"**: Exit plan mode.
-
-After artifact creation, exit plan mode and confirm:
-
-```
-Created [ARTIFACT] from synthesis "[TOPIC]"
-
-**Multi-Expert Analysis Summary:**
-- Experts consulted: {list}
-- Consensus level: {High/Medium/Low}
-- Key insights preserved: {count}
-
-**Artifacts:**
-- Synthesis: docs/10-research/YYYYMMDD-synthesis-topic.md
-- [Artifact]: [path or ID]
-```
-
----
-
-## Saving Synthesis Report
-
-If user chooses to save:
-
-**Filename format**: `YYYYMMDD-synthesis-[topic-slug].md`
-
-**Add to docs/10-research/README.md index (5-column format):**
-```markdown
-| Date | Topic | Type | Path | Summary |
-|------|-------|------|------|---------|
-| 2026-01-17 | Synthesis: Authentication | Synthesis | [20260117-synthesis-authentication.md](./20260117-synthesis-authentication.md) | Cross-research synthesis of N files... |
-```
-
-**Mark as synthesis type** in file header:
-```markdown
-# Synthesis: [Topic]
-
-**Type**: Synthesis (cross-research analysis)
-**Date**: YYYY-MM-DD
-**Files Analyzed**: N
-**Consensus Items**: N
-**Conflicts Found**: N
-```
-
----
-
-## Rules
-
-- **Read actual file contents** - Don't guess based on titles
-- **Apply confidence scoring** - HIGH/UNIQUE/CONFLICT
-- **Flag conflicts explicitly** - They need human review
-- **Show synthesis before artifacts** - Diff-first principle
-- **Link to source files** - Always reference original research
-- **Age-aware analysis** - Older research may be outdated
-
----
-
-## Related Commands
-
-- `/agileflow:research:list` - Show all research notes
-- `/agileflow:research:view` - Read specific research note
-- `/agileflow:research:analyze` - Analyze single research for implementation
-- `/agileflow:research:import` - Import new research
-- `/agileflow:research:ask` - Generate research prompt for web AI