@su-record/vibe 2.5.19 → 2.5.20

package/CLAUDE.md

@@ -40,14 +40,17 @@ Follow these standards when writing code. See `~/.claude/vibe/rules/` (global) f
 ## Workflow
 
 ```
-/vibe.spec → (auto) SPEC review → /vibe.run → (auto) code review → (auto) fix → ✅ Done
+/vibe.spec → /new → /vibe.spec.review → /vibe.run → (auto) code review → ✅ Done
 ```
 
-**Automated Flow:**
-1. `/vibe.spec` - Write SPEC + **(auto)** Gemini review → Auto-apply
-2. `/vibe.run` - Implementation + Gemini review
-3. **(auto)** 13+ agent parallel review
-4. **(auto)** P1/P2 issue auto-fix
+**Flow:**
+
+1. `/vibe.spec` - Write SPEC (requirements + research + draft)
+2. `/new` - Start new session (clean context)
+3. `/vibe.spec.review` - GPT/Gemini review (3-round mandatory)
+4. `/vibe.run` - Implementation + Gemini review
+5. **(auto)** 13+ agent parallel review
+6. **(auto)** P1/P2 issue auto-fix
 
 ## Plan Mode vs VIBE (Workflow Selection)
 
@@ -102,6 +105,7 @@ Include `ultrawork` or `ulw` keyword to activate maximum performance mode:
 | Command | Description |
 |---------|-------------|
 | `/vibe.spec "feature-name"` | Write SPEC (PTCF structure) + parallel research |
+| `/vibe.spec.review "feature-name"` | **GPT/Gemini review** (run in new session) |
 | `/vibe.run "feature-name"` | Execute implementation |
 | `/vibe.run "feature-name" ultrawork` | **Maximum performance mode** |
 | `/vibe.verify "feature-name"` | Verification |

package/commands/vibe.spec.md

@@ -280,8 +280,11 @@ Read ~/.claude/vibe/languages/typescript-react.md
 
 **→ IMMEDIATELY run these 4 Bash commands IN PARALLEL (all at once):**
 
+**🚨 MANDATORY: Copy the EXACT path below. DO NOT modify or use alternative paths.**
+
 ```bash
 # Cross-platform path (works on Windows/macOS/Linux)
+# ⚠️ COPY THIS EXACTLY - DO NOT USE ~/.claude/ or any other path!
 VIBE_SCRIPTS="$(node -p "process.env.APPDATA || require('os').homedir() + '/.config'")/vibe/hooks/scripts"
 
 # 1. GPT: Best practices
@@ -705,7 +708,7 @@ Grades:
 
 #### 7.3 Quality Gate (Auto-verification)
 
-**Minimum 95 points required to proceed to next step (GPT+Gemini review)**
+**Minimum 95 points required to complete SPEC draft**
 
 ```
 SPEC writing complete
@@ -714,7 +717,7 @@ SPEC writing complete
       ↓
 Score < 95? → Show missing items → Attempt auto-fix → Re-evaluate
       ↓
-Score ≥ 95 → Proceed to GPT + Gemini review
+Score ≥ 95 → SPEC Draft Complete → Handoff to /vibe.spec.review
 ```
 
 #### 7.4 Auto-Fix for Low Score
@@ -728,182 +731,33 @@ If score is below 95, attempt automatic fixes:
 | Missing error handling | Add common error scenarios |
 | Missing performance targets | Apply industry standard criteria |
 
-### 8. SPEC Review (GPT/Gemini) - 3-Round Mandatory Review
-
-**🚨 CRITICAL: YOU MUST EXECUTE ALL 3 ROUNDS. DO NOT SKIP.**
-
-```bash
-/vibe.spec "feature"          # Default: 3 rounds (MANDATORY)
-/vibe.spec "feature" --quick  # Quick mode: 1 round only
-```
-
-**3-Round Review Process (MANDATORY):**
-
-```
-# 🚨 CRITICAL: This is a FOR loop, NOT a WHILE loop with early exit
-# You MUST complete all 3 rounds regardless of feedback
-
-FOR round_count IN [1, 2, 3]:  # EXACTLY 3 rounds, NO early exit
-
-    print(f"━━━ Round {round_count}/3 ━━━")
-
-    # 1. Call GPT + Gemini in parallel (Bash)
-    gpt_result = Bash("node llm-orchestrate.js gpt ...")
-    gemini_result = Bash("node llm-orchestrate.js gemini ...")
-
-    # 2. Merge and apply feedback (even if "no issues", still run next round)
-    apply_feedback(gpt_result, gemini_result)
-    update_spec()
-
-    # 3. Show progress
-    print(f"✅ Round {round_count}/3 complete")
-
-    # 4. NO EARLY EXIT - continue to next round
-    # Even if both say "no issues", proceed to next round for verification
+### 8. SPEC Draft Complete - Handoff to Review
 
-# After ALL 3 rounds complete:
-print(f"🎉 Review complete! All 3 rounds finished.")
-```
-
-**🚨 ABSOLUTE RULES (NO EXCEPTIONS):**
-1. **MUST complete Round 1** - never skip
-2. **MUST complete Round 2** - even if Round 1 had no issues
-3. **MUST complete Round 3** - even if Round 1 and 2 had no issues
-4. **MUST show "Round X/3"** in output for EACH round
-5. **NO EARLY EXIT** - all 3 rounds are mandatory for quality assurance
-6. **VIOLATION = SPEC NOT COMPLETE** - if less than 3 rounds, SPEC is invalid
+**🚨 IMPORTANT: GPT/Gemini review is now a SEPARATE command**
 
-**Output format for each round:**
+After SPEC draft is complete (score ≥ 95), output the handoff message:
 
 ```
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-🔍 SPEC REVIEW - Round 1/3
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-[GPT] 3 issues found: ...
-[Gemini] 2 issues found: ...
-✅ Applied 5 improvements
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-🔍 SPEC REVIEW - Round 2/3
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-[GPT] 1 issue found: ...
-[Gemini] 0 issues found
-✅ Applied 1 improvement
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+✅ SPEC DRAFT COMPLETE: {feature-name}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-🔍 SPEC REVIEW - Round 3/3
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-[GPT] 0 issues found
-[Gemini] 0 issues found
-✅ No changes needed
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-🎉 SPEC Review complete! 3 rounds, 6 total improvements
-```
+📋 SPEC: .claude/vibe/specs/{feature-name}.spec.md
+📋 Feature: .claude/vibe/features/{feature-name}.feature
+📊 Quality Score: {score}/100
 
-**After SPEC draft, execute review loop:**
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+⚠️ NEXT STEP: Run GPT/Gemini review in a NEW session
 
+1. Start new session: /new
+2. Run review: /vibe.spec.review "{feature-name}"
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 ```
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-🔍 SPEC REVIEW
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-[Step 1] Sending SPEC to Gemini/GPT...
-  - Full SPEC content
-  - Feature file (scenarios)
-  - Project context
-
-[Step 2] Review feedback:
-  ┌─────────────────────────────────────────┐
-  │ 📝 SPEC Review Feedback                 │
-  │                                         │
-  │ 1. [Missing] Error handling scenarios   │
-  │    → Recommend adding "retry on network │
-  │      error"                             │
-  │                                         │
-  │ 2. [Security] Auth token expiry         │
-  │    handling undefined                   │
-  │    → Recommend adding refresh token     │
-  │      flow                               │
-  │                                         │
-  │ 3. [Edge] Concurrent login policy       │
-  │    undefined                            │
-  │    → Need to specify existing session   │
-  │      handling method                    │
-  └─────────────────────────────────────────┘
-
-[Step 3] Auto-applying...
-  ✅ Added error handling Phase to SPEC Task
-  ✅ Added token expiry scenario to Feature
-  ✅ Added concurrent login policy to Constraints
-
-[Step 4] Re-verifying...
-  ✅ Ambiguity Scan: 0 issues
-  ✅ Quality Score: 95/100
-
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-✅ SPEC Review complete! 3 improvements applied
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-```
-
-**MUST: SPEC Review (Required) - GPT + Gemini Parallel Execution**
-
-🚨 **Run GPT and Gemini in parallel for cross-validation to ensure quality**
-
-**Parallel execution (2 Bash calls simultaneously):**
-
-```bash
-# Cross-platform path (works on Windows/macOS/Linux)
-VIBE_SCRIPTS="$(node -p "process.env.APPDATA || require('os').homedir() + '/.config'")/vibe/hooks/scripts"
-
-# GPT review (code patterns, architecture perspective)
-node "$VIBE_SCRIPTS/llm-orchestrate.js" gpt orchestrate-json "Review SPEC for [feature-name]. Stack: [stack]. Summary: [summary]. Check: completeness, error handling, security, edge cases."
-
-# Gemini review (latest docs, best practices perspective)
-node "$VIBE_SCRIPTS/llm-orchestrate.js" gemini orchestrate-json "Review SPEC for [feature-name]. Stack: [stack]. Summary: [summary]. Check: completeness, error handling, security, edge cases."
-```
-
-**Important:**
-
-- **Must call both GPT and Gemini** (parallel execution)
-- Merge feedback from both models and apply to SPEC
-- If one model fails, apply only the remaining result
-- Don't send full SPEC, send only **key summary** (save tokens)
-- Parse JSON response and auto-apply to SPEC
-
-**Why parallel execution:**
-
-| Model | Strengths |
-|-------|-----------|
-| GPT | Code patterns, architecture, logic validation |
-| Gemini | Latest web info, doc search, best practices |
-
-**Result merge:** Combine feedback from both models for more thorough review
-
-**Review items:**
-
-| Category | Check Point |
-|----------|-------------|
-| Completeness | All user flows covered? |
-| Error Handling | Failure scenarios defined? |
-| Security | Auth/authorization/data protection? |
-| Edge Cases | Boundary conditions handled? |
-| Testability | ACs verifiable? |
-
-**Auto-apply rules:**
-
-| Feedback Type | Action |
-|---------------|--------|
-| Missing scenario | Auto-add to Feature |
-| Security considerations | Auto-add to Constraints |
-| Edge cases | Auto-add to Task Phase |
-| Lack of clarity | Supplement the section |
 
-**Fallback handling:**
-- On `"status": "fallback"` response → Skip and proceed to next step
-- On network error → Retry once then skip
+**Why separate session?**
+- Prevents context bloat from affecting review accuracy
+- Ensures GPT/Gemini paths and instructions are followed correctly
+- Review command is short and focused
 
 ## Output (MANDATORY File Creation)
 

package/commands/vibe.spec.review.md

@@ -0,0 +1,300 @@
+# /vibe.spec.review - SPEC Quality Review
+
+Review and enhance SPEC with GPT/Gemini cross-validation.
+
+**Purpose:** Run this command in a NEW session after `/vibe.spec` to ensure accurate review execution.
+
+---
+
+## Usage
+
+```bash
+/vibe.spec.review "feature-name"
+```
+
+**Prerequisites:**
+- SPEC file exists: `.claude/vibe/specs/{feature-name}.spec.md`
+- Feature file exists: `.claude/vibe/features/{feature-name}.feature`
+
+---
+
+## Workflow
+
+```
+/vibe.spec "feature" → SPEC 생성 완료
+        ↓
+    /new (새 세션)
+        ↓
+/vibe.spec.review "feature" → 품질 검증 + GPT/Gemini 리뷰
+        ↓
+    /vibe.run "feature"
+```
+
+---
+
+## Step 1: Load SPEC Files
+
+Read the existing SPEC and Feature files:
+
+```
+.claude/vibe/specs/{feature-name}.spec.md
+.claude/vibe/features/{feature-name}.feature
+```
+
+**Output:**
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+📋 SPEC REVIEW: {feature-name}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+Loading files...
+  ✅ SPEC: .claude/vibe/specs/{feature-name}.spec.md
+  ✅ Feature: .claude/vibe/features/{feature-name}.feature
+
+Extracted info:
+  - Feature: {feature description}
+  - Stack: {tech stack}
+  - Phases: {number of phases}
+  - Scenarios: {number of scenarios}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+---
+
+## Step 2: Quality Validation (95-Point Gate)
+
+**🚨 MANDATORY: Score must be ≥ 95 to proceed**
+
+### 2.1 Quality Checklist
+
+| Category | Check Item | Weight |
+|----------|------------|--------|
+| **Completeness** | All user flows included in Task | 15% |
+| **Completeness** | All ACs converted to Feature scenarios | 10% |
+| **Completeness** | Error handling scenarios defined | 10% |
+| **Specificity** | All numbers specified (timeout, limits, etc.) | 15% |
+| **Specificity** | No ambiguous terms ("appropriate", "proper", etc.) | 10% |
+| **Testability** | Each AC is verifiable | 10% |
+| **Testability** | Feature scenarios have concrete Given/When/Then | 10% |
+| **Security** | Auth/permission requirements specified | 10% |
+| **Performance** | Response time/load requirements specified | 10% |
+
+### 2.2 Quality Gate Loop
+
+```python
+max_iterations = 3
+iteration = 0
+
+while iteration < max_iterations:
+    iteration += 1
+    score = calculate_quality_score(spec, feature)
+
+    print(f"━━━ Quality Check [{iteration}/{max_iterations}] ━━━")
+    print(f"Score: {score}/100")
+
+    if score >= 95:
+        print("✅ Quality Gate PASSED")
+        break
+
+    # Auto-fix missing items
+    missing_items = identify_missing_items(spec)
+    for item in missing_items:
+        auto_fix(item)
+        update_spec()
+        update_feature()
+
+    print(f"✅ Applied {len(missing_items)} fixes - Re-evaluating...")
+
+if score < 95:
+    print(f"❌ BLOCKED: Score {score} < 95 after {max_iterations} iterations")
+    print("Manual intervention required.")
+```
+
+**Output format:**
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+📊 QUALITY GATE [1/3]
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+Score: 87/100 ⚠️ BELOW THRESHOLD (95)
+
+Missing items:
+  ❌ Error handling scenarios (10%)
+  ❌ Performance targets (5%)
+
+Auto-fixing...
+  ✅ Added network error handling scenario
+  ✅ Added response time targets (<500ms)
+
+Re-evaluating...
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+📊 QUALITY GATE [2/3]
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+Score: 96/100 ✅ PASSED
+
+✅ Quality Gate PASSED - proceeding to GPT/Gemini review
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+### 2.3 Auto-Fix Rules
+
+| Missing Item | Auto-Fix Method |
+|--------------|-----------------|
+| Missing AC | Auto-generate AC based on Task |
+| Numbers not specified | Apply project defaults (timeout 30s, etc.) |
+| Missing error handling | Add common error scenarios |
+| Missing performance targets | Apply industry standard criteria |
+| Missing security | Add auth/data protection requirements |
+| Ambiguous terms | Replace with specific values |
+
+---
+
+## Step 3: GPT/Gemini Review (3-Round Mandatory)
+
+**🚨 CRITICAL: Execute ALL 3 rounds. DO NOT skip.**
+
+### 3.1 Path Configuration
+
+**🚨 MANDATORY: Copy the EXACT path below. DO NOT modify or use alternative paths.**
+
+```bash
+# Cross-platform path (works on Windows/macOS/Linux)
+# ⚠️ COPY THIS EXACTLY - DO NOT USE ~/.claude/ or any other path!
+VIBE_SCRIPTS="$(node -p "process.env.APPDATA || require('os').homedir() + '/.config'")/vibe/hooks/scripts"
+```
+
+### 3.2 Review Loop (3 Rounds)
+
+For each round (1 to 3):
+
+**Run GPT and Gemini in PARALLEL (2 Bash calls simultaneously):**
+
+```bash
+# GPT review
+node "$VIBE_SCRIPTS/llm-orchestrate.js" gpt orchestrate-json "Review SPEC for {feature-name}. Stack: {stack}. Summary: {spec-summary}. Round {N}/3. Check: completeness, error handling, security, edge cases, performance."
+
+# Gemini review
+node "$VIBE_SCRIPTS/llm-orchestrate.js" gemini orchestrate-json "Review SPEC for {feature-name}. Stack: {stack}. Summary: {spec-summary}. Round {N}/3. Check: completeness, error handling, security, edge cases, performance."
+```
+
+**After each round:**
+1. Parse JSON responses from both GPT and Gemini
+2. Merge feedback (deduplicate, prioritize)
+3. Auto-apply improvements to SPEC and Feature files
+4. Continue to next round
+
+**Output format:**
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+🔍 SPEC REVIEW - Round 1/3
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+[GPT] 2 issues found:
+  1. Missing retry logic for API calls
+  2. Token refresh flow not specified
+
+[Gemini] 1 issue found:
+  1. Missing rate limiting specification
+
+Merged: 3 unique issues
+Auto-applying...
+  ✅ Added retry logic (3 attempts, exponential backoff)
+  ✅ Added token refresh flow to auth section
+  ✅ Added rate limiting (100 req/min)
+
+✅ Round 1 complete - 3 improvements applied
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+🔍 SPEC REVIEW - Round 2/3
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+[GPT] 1 issue found:
+  1. Concurrent session handling unclear
+
+[Gemini] 0 issues found
+
+Auto-applying...
+  ✅ Added concurrent session policy
+
+✅ Round 2 complete - 1 improvement applied
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+🔍 SPEC REVIEW - Round 3/3
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+[GPT] 0 issues found
+[Gemini] 0 issues found
+
+✅ No changes needed - SPEC is complete
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+---
+
+## Step 4: Final Summary
+
+```
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+✅ SPEC REVIEW COMPLETE: {feature-name}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+Quality Score: 96/100 ✅
+Review Rounds: 3/3 ✅
+Total Improvements: 4
+
+Updated files:
+  📋 .claude/vibe/specs/{feature-name}.spec.md
+  📋 .claude/vibe/features/{feature-name}.feature
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+## Next Step
+
+/vibe.run "{feature-name}"
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+---
+
+## Error Handling
+
+### SPEC Not Found
+```
+❌ ERROR: SPEC file not found
+
+Expected: .claude/vibe/specs/{feature-name}.spec.md
+
+Please run /vibe.spec "{feature-name}" first to create the SPEC.
+```
+
+### Feature Not Found
+```
+❌ ERROR: Feature file not found
+
+Expected: .claude/vibe/features/{feature-name}.feature
+
+Please run /vibe.spec "{feature-name}" first to create the Feature file.
+```
+
+### GPT/Gemini Call Failed
+```
+⚠️ WARNING: {GPT|Gemini} call failed
+
+Error: {error message}
+
+Continuing with {other model} results only...
+```
+
+---
+
+## Quick Mode
+
+For faster iteration (1 round only):
+
+```bash
+/vibe.spec.review "feature-name" --quick
+```
+
+---
+
+ARGUMENTS: $ARGUMENTS

package/hooks/scripts/generate-brand-assets.js

@@ -52,7 +52,9 @@ function getGeminiApiKey() {
 }
 
 async function generateImageWithGemini(prompt, apiKey) {
-  const url = `https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp:generateContent?key=${apiKey}`;
+  // Nano Banana (Gemini 2.5 Flash Image) - fast image generation for icons/logos
+  // For professional assets, use gemini-3-pro-image-preview (Nano Banana Pro)
+  const url = `https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent?key=${apiKey}`;
 
   const requestBody = {
     contents: [{

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@su-record/vibe",
-  "version": "2.5.19",
+  "version": "2.5.20",
   "description": "Vibe - Claude Code exclusive SPEC-driven AI coding framework with 35+ integrated tools",
   "type": "module",
   "main": "dist/cli/index.js",