cmp-standards 3.8.1 → 3.9.0

package/templates/commands/experts.md

@@ -1,6 +1,6 @@
 ---
 name: experts
-description: Multi-expert code review with rigorous logical reasoning and evidence-based consensus
+description: Multi-expert code review with rigorous logical reasoning, evidence-based consensus, and autonomous interaction style selection
 arguments:
   - name: target
     description: Files, directories, or task description to review
@@ -20,11 +20,99 @@ arguments:
 # /experts - Evidence-Based Code Review System
 
 > **Framework**: All experts use `_reasoning-framework.md` for rigorous logical analysis.
+> **Version**: 3.0.0 (with Autonomous Deliberation)
 
 Execute this multi-phase expert review following the logical sequence below.
 
 ---
 
+## PHASE 0: AUTONOMOUS STYLE SELECTION (NEW)
+
+Before executing, the orchestrator autonomously selects the optimal interaction style.
+
+### Interaction Styles
+
+| Style | When Used | Expert Behavior |
+|-------|-----------|-----------------|
+| **SUMMARY** | Simple reviews (≤3 files, low complexity) | Quick cards, expand on demand |
+| **DEBATE** | Trade-offs detected, experts likely to disagree | Experts discuss, user decides |
+| **SOCRATIC** | Complex new features, learning opportunity | Questions instead of findings |
+| **CLASSIC** | Critical code (auth, payment, schema) | Full rigorous analysis |
+
+### Style Selection Algorithm
+
+```typescript
+function selectInteractionStyle(ctx: ReviewContext): InteractionStyle {
+  // CRITICAL code ALWAYS gets full analysis - no shortcuts
+  if (ctx.isCritical) return 'CLASSIC'
+
+  // Trade-offs detected → Let experts debate
+  if (ctx.hasTradeoffs || ctx.hasConflicts) return 'DEBATE'
+
+  // Complex new features → Guide with questions
+  if (ctx.changeType === 'new_feature' && ctx.complexity === 'high') return 'SOCRATIC'
+
+  // Simple reviews → Quick summaries
+  if (ctx.fileCount <= 3 && ctx.complexity === 'low') return 'SUMMARY'
+
+  // Default → Summary with expansion capability
+  return 'SUMMARY'
+}
+```
+
+### Trade-off Detection Patterns
+
+```typescript
+const TRADEOFF_SIGNALS = [
+  // Performance vs Readability
+  /useMemo|useCallback|React\.memo/,
+
+  // Security vs UX
+  /rate.?limit|captcha|2fa|mfa/,
+
+  // Flexibility vs Complexity
+  /generic|<T>|abstract|interface.*extends/,
+
+  // Consistency vs Speed
+  /cache|memoize|stale.?while/,
+
+  // DRY vs Explicit
+  /helper|util|shared|common/,
+]
+```
+
+### Transparency Announcement
+
+Before executing, ALWAYS announce the chosen approach:
+
+```
+┌──────────────────────────────────────────────────────────────────┐
+│ 📋 REVIEW APPROACH                                               │
+├──────────────────────────────────────────────────────────────────┤
+│ Files: {count} files, ~{lines} lines                             │
+│ Domains: {domains}                                               │
+│ Criticality: {CRITICAL|NORMAL}                                   │
+│                                                                  │
+│ {Reasoning for chosen style}                                     │
+│                                                                  │
+│ Style: {SUMMARY|DEBATE|SOCRATIC|CLASSIC}                         │
+│ Experts: {list}                                                  │
+└──────────────────────────────────────────────────────────────────┘
+```
+
+### Adaptive Triggers
+
+During the review, switch styles if:
+
+| Trigger | New Style | Example |
+|---------|-----------|---------|
+| User asks "why?" or "explain" | Expand current expert | → Full analysis |
+| Experts disagree | DEBATE | Security APPROVE, Perf REJECT |
+| User seems confused | SOCRATIC | → Guide with questions |
+| User says "just tell me" | SUMMARY | → Condensed output |
+
+---
+
 ## PHASE 1: PRE-ANALYSIS (Mandatory First Step)
 
 Before invoking ANY expert, complete these checks:
@@ -99,27 +187,68 @@ Launch selected experts in parallel using the Task tool.
 
 ### 2.1 Expert Invocation Template
 
-For EACH selected expert, invoke with this prompt structure:
+For EACH selected expert, invoke with this **style-aware** prompt structure:
 
 ```
-You are the {EXPERT_NAME} reviewing these files: {FILE_LIST}
+[EXPERT REVIEW]
+Files: {FILE_LIST}
+Criticality: {CRITICAL|NORMAL}
+Context: {1-line description}
+Change Type: {feature|bugfix|refactor}
+Style: {SUMMARY|DEBATE|SOCRATIC|CLASSIC}
 
 MANDATORY: Follow the reasoning framework in _reasoning-framework.md
 
-FOR EACH FINDING:
-1. OBSERVE: Quote exact code with file:line
-2. STATE PREMISE: "IF [condition] THEN [consequence]"
-3. VERIFY: Show what you searched and found
-4. FALSIFY: Ask "What would make this NOT an issue?"
-5. CONCLUDE: State confidence level and severity
+{STYLE-SPECIFIC INSTRUCTIONS}
+[END REVIEW]
+```
+
+### Style-Specific Instructions
 
-CONFIDENCE REQUIREMENTS:
-- CERTAIN: Direct observation, reproducible
-- HIGH: Pattern match + context verified
-- MEDIUM: Pattern match, some gaps
-- LOW: Inference only (cannot trigger REJECT)
+| Style | Instructions Added to Prompt |
+|-------|------------------------------|
+| **SUMMARY** | "Provide a 2-3 sentence summary and your single most important finding. Save detailed analysis for expansion." |
+| **DEBATE** | "State your position clearly. Other experts may disagree. Include reasoning for trade-offs. Be prepared to respond to counter-arguments." |
+| **SOCRATIC** | "Instead of findings, generate 2-3 thought-provoking questions. Categories: clarify, challenge, explore, validate. Guide the developer to discover issues." |
+| **CLASSIC** | "Full analysis with OBSERVE → PREMISE → VERIFY → FALSIFY → CONCLUDE. No shortcuts." |
+
+### Expected Output Per Style
+
+```typescript
+// SUMMARY mode output
+{
+  "expert": "{EXPERT_NAME}",
+  "vote": "APPROVE|REJECT|ABSTAIN",
+  "summary": "2-3 sentence summary",
+  "keyFinding": "Single most important finding",
+  "expandable": true,
+  "fullAnalysis": "...saved for expansion..."
+}
+
+// DEBATE mode output
+{
+  "expert": "{EXPERT_NAME}",
+  "vote": "APPROVE|REJECT|ABSTAIN",
+  "position": "Your stance on the approach",
+  "stance": "agrees|disagrees|expands|questions",
+  "stanceTarget": "which expert (if responding)",
+  "reasoning": "Why you hold this position",
+  "tradeoff": "Performance vs Readability / Security vs UX etc"
+}
+
+// SOCRATIC mode output
+{
+  "expert": "{EXPERT_NAME}",
+  "questions": [
+    {
+      "question": "What happens if...?",
+      "purpose": "challenge|clarify|explore|validate",
+      "options": ["Option A", "Option B", "Option C"]
+    }
+  ]
+}
 
-OUTPUT FORMAT (JSON):
+// CLASSIC mode output (original)
 {
   "expert": "{EXPERT_NAME}",
   "vote": "APPROVE|REJECT|ABSTAIN",