@su-record/vibe 2.4.27 → 2.4.29

package/agents/searcher.md

@@ -11,24 +11,24 @@ Web search specialist sub-agent.
 
 ## Model
 
-- **With GPT integration**: Use GPT (mcp__vibe-gpt__search)
 - **Default**: Haiku 4.5 + WebSearch
+- **Fallback**: GPT hook (`gpt- [query]`)
 
 ## Usage
 
 ```
-# With GPT integration
-mcp__vibe-gpt__search("React 19 changes")
-
 # Default (Haiku + WebSearch)
 Task(model: "haiku", prompt: "Search React 19 changes")
+
+# Fallback to GPT
+gpt- Search for React 19 changes and new features
 ```
 
 ## Tools
 
 - WebSearch - Web search (default)
 - WebFetch - Fetch page content
-- mcp__vibe-gpt__* - GPT search (when integrated)
+- GPT hook - `gpt- [query]` (fallback)
 
 ## Process
 

package/commands/vibe.run.md

@@ -267,29 +267,20 @@ Automatically select optimal model based on task type:
 
 When external LLMs are enabled in `.claude/vibe/config.json`:
 
-| Role | Model | Condition |
-|------|-------|-----------|
-| Architecture/Debugging | GPT 5.2 | When `vibe gpt <key>` executed |
-| UI/UX Design, Exploration | Gemini 2.5/3 | When `vibe gemini --auth` executed |
-
-When external LLM enabled, automatically called via MCP:
-- `mcp__vibe-gpt__chat` - GPT 5.2 architecture consultation
-- `mcp__vibe-gemini__gemini_chat` - Gemini 질문/상담
-- `mcp__vibe-gemini__gemini_analyze_code` - 코드 분석
-- `mcp__vibe-gemini__gemini_review_ui` - UI/UX 리뷰
-- `mcp__vibe-gemini__gemini_quick_ask` - 빠른 질문 (탐색용)
+| Role | Prefix | Condition |
+|------|--------|-----------|
+| Architecture/Debugging | `gpt-` or `지피티-` | When `vibe gpt login` executed |
+| UI/UX Design, Exploration | `gemini-` or `제미나이-` | When `vibe gemini login` executed |
 
-### External LLM Fallback
+When external LLM enabled, use hook prefixes:
+- `gpt- [question]` - GPT architecture consultation
+- `gemini- [question]` - Gemini 질문/상담
+- `gemini- Analyze this code: [code]` - 코드 분석
+- `gemini- Review UI/UX for: [component]` - UI/UX 리뷰
 
-**IMPORTANT**: When Gemini/GPT MCP returns `"status": "fallback"`, Claude MUST handle the task directly:
+### External LLM Fallback
 
-```json
-{
-  "status": "fallback",
-  "reason": "rate_limit",  // or "auth_error"
-  "message": "Gemini API 할당량 초과. Claude가 직접 처리해주세요."
-}
-```
+**IMPORTANT**: When GPT/Gemini hook fails, Claude MUST handle the task directly:
 
 **Fallback behavior**:
 - Do NOT retry the external LLM call
@@ -449,8 +440,8 @@ Then: 로그인 성공 + JWT 토큰 반환
 │               │                                                 │
 │  Task(haiku) ─┴─→ "Find existing patterns and conventions"      │
 │                                                                 │
-│  [If GPT enabled] + MCP(vibe-gpt): Architecture review          │
-│  [If Gemini enabled] + MCP(vibe-gemini): UI/UX consultation     │
+│  [If GPT enabled] gpt- Review architecture: [design]            │
+│  [If Gemini enabled] gemini- Suggest UX for: [component]        │
 └─────────────────────────────────────────────────────────────────┘
                               │
                               ↓ (wait for all to complete)
@@ -655,22 +646,27 @@ Phase N+1 Start (IMMEDIATE - exploration already done!)
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 ```
 
-**MUST: Gemini MCP 호출 (필수)**
+**MUST: Gemini 코드 리뷰 (필수)**
 
-Gemini MCP가 활성화된 경우, **반드시** 아래 MCP를 호출하여 코드 리뷰를 받아야 합니다:
+Gemini가 활성화된 경우, **반드시** 아래 훅을 사용하여 코드 리뷰를 받아야 합니다:
+
+```
+gemini- Review this code for security, performance, best-practices:
+[변경된 파일들의 전체 코드]
+SPEC 요구사항: [요약]
+시나리오: [구현한 시나리오 목록]
+```
 
+또는 한글로:
 ```
-mcp__vibe-gemini__gemini_analyze_code({
-  code: "[변경된 파일들의 전체 코드]",
-  context: "SPEC 요구사항: [요약]\n시나리오: [구현한 시나리오 목록]",
-  focus: "security,performance,best-practices"
-})
+제미나이- 이 코드 리뷰해줘 (보안, 성능, 베스트프랙티스):
+[코드]
 ```
 
 **호출 순서:**
-1. 변경된 모든 파일 내용을 `code` 파라미터에 포함
-2. SPEC의 핵심 요구사항을 `context`에 요약
-3. MCP 호출 실행
+1. 변경된 모든 파일 내용을 질문에 포함
+2. SPEC의 핵심 요구사항 요약 추가
+3. 훅 호출 실행
 4. 응답의 피드백 항목별로 코드 수정
 5. 빌드/테스트 재실행
 

package/commands/vibe.spec.md

@@ -28,30 +28,30 @@ Collect requirements through conversation with the user and create an **AI-execu
 
 ## External LLM Integration (Optional)
 
-When external LLMs are enabled in `.claude/vibe/config.json`, automatically utilize during SPEC creation:
+When external LLMs are enabled, automatically utilize during SPEC creation:
 
 ```
 /vibe.spec "complex feature"
       ↓
 [Claude Opus] Create SPEC draft
       ↓
-[GPT enabled?] → Cross-review design via MCP(vibe-gpt)
+[GPT enabled?] → gpt- Review this architecture: [design]
       ↓
-[Gemini enabled?] → UI/UX consultation via MCP(vibe-gemini)
+[Gemini enabled?] → gemini- Suggest UX improvements for: [component]
       ↓
 [Claude] Finalize SPEC
 ```
 
-| External LLM | Role | When Used |
-|--------------|------|-----------|
-| GPT 5.2 | Architecture/design review | After SPEC draft completion |
-| Gemini 3 | UI/UX consultation | During design reference discussion |
+| External LLM | Prefix | Role | When Used |
+|--------------|--------|------|-----------|
+| GPT | `gpt-` or `지피티-` | Architecture/design review | After SPEC draft completion |
+| Gemini | `gemini-` or `제미나이-` | UI/UX consultation | During design reference discussion |
 
 **Activation:**
 ```bash
-vibe gpt <api-key>      # Enable GPT
-vibe gemini <api-key>   # Enable Gemini
-vibe status             # Check current settings
+vibe gpt login      # Enable GPT (OAuth)
+vibe gemini login   # Enable Gemini (OAuth)
+vibe status         # Check current settings
 ```
 
 ## Process

package/hooks/hooks.json

@@ -122,21 +122,21 @@
         ]
       },
       {
-        "matcher": "(?!.*gemini)(gpt\\s*한테|gpt\\s*에게|gpt에\\s*물어|gpt.*질문|ask\\s*gpt)",
+        "matcher": "^(gpt-|지피티-)",
         "hooks": [
           {
             "type": "command",
-            "command": "node -e \"const V='file:///{{VIBE_PATH}}/node_modules/@su-record/vibe/dist/lib/gpt-api.js';let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{const q=JSON.parse(d).prompt.replace(/gpt.*?(물어|질문|한테|분석)[^가-힣]*/i,'').trim();import(V).then(g=>g.quickAsk(q)).then(r=>console.log('[GPT] '+r)).catch(e=>console.log('[GPT] Error:',e.message))})\"",
+            "command": "node -e \"const V='file:///{{VIBE_PATH}}/node_modules/@su-record/vibe/dist/lib/gpt-api.js';let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{const q=JSON.parse(d).prompt.replace(/^(gpt-|지피티-)\\s*/i,'').trim();import(V).then(g=>g.quickAsk(q)).then(r=>console.log('[GPT] '+r)).catch(e=>console.log('[GPT] Error:',e.message))})\"",
             "statusMessage": "Asking GPT..."
           }
         ]
       },
       {
-        "matcher": "(?!.*gpt)(gemini\\s*한테|gemini\\s*에게|gemini에\\s*물어|gemini.*질문|ask\\s*gemini|제미나이)",
+        "matcher": "^(gemini-|제미나이-)",
         "hooks": [
           {
             "type": "command",
-            "command": "node -e \"const V='file:///{{VIBE_PATH}}/node_modules/@su-record/vibe/dist/lib/gemini-api.js';let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{const q=JSON.parse(d).prompt.replace(/(gemini|제미나이).*?(물어|질문|한테|분석)[^가-힣]*/i,'').trim();import(V).then(g=>g.quickAsk(q)).then(r=>console.log('[Gemini] '+r)).catch(e=>console.log('[Gemini] Error:',e.message))})\"",
+            "command": "node -e \"const V='file:///{{VIBE_PATH}}/node_modules/@su-record/vibe/dist/lib/gemini-api.js';let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{const q=JSON.parse(d).prompt.replace(/^(gemini-|제미나이-)\\s*/i,'').trim();import(V).then(g=>g.quickAsk(q)).then(r=>console.log('[Gemini] '+r)).catch(e=>console.log('[Gemini] Error:',e.message))})\"",
             "statusMessage": "Asking Gemini..."
           }
         ]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@su-record/vibe",
-  "version": "2.4.27",
+  "version": "2.4.29",
   "description": "Vibe - Claude Code exclusive SPEC-driven AI coding framework with 35+ integrated tools",
   "type": "module",
   "main": "dist/cli/index.js",

package/skills/context7-usage.md

@@ -78,5 +78,5 @@ context7 fails
     ↓
 Web Search for official docs
     ↓
-Ask vibe-gpt
+gpt- [library question]
 ```

package/skills/multi-llm-orchestration.md

@@ -3,50 +3,45 @@ description: Multi-LLM collaboration guide. Auto-activates for architecture revi
 ---
 # Multi-LLM Orchestration
 
-Guide for using GPT/Gemini as sub-agents in Claude Code.
+Guide for using GPT/Gemini as sub-agents in Claude Code via hooks.
 
-## When to Use GPT (mcp__vibe-gpt)
+## Calling GPT
 
-| Situation | Reason |
-|-----------|--------|
-| Architecture design/review | Get different perspective on design |
-| Complex debugging | Fresh eyes on the problem |
-| Algorithm optimization | Alternative algorithm suggestions |
-| Security review | Cross-check vulnerabilities |
-| Technology decisions | Compare pros/cons |
+Use `gpt-` or `지피티-` prefix:
 
-### Usage
+| Situation | Example |
+|-----------|---------|
+| Architecture review | `gpt- Review this auth architecture: [code]` |
+| Complex debugging | `gpt- Debug this error: [error message]` |
+| Algorithm optimization | `gpt- Suggest better algorithm for: [problem]` |
+| Security review | `gpt- Check security vulnerabilities in: [code]` |
+| Technology decisions | `gpt- Compare Redis vs Memcached for caching` |
+
+### Usage Examples
 
 ```
-mcp__vibe-gpt__gpt_chat({
-  prompt: "Review this auth architecture: [code/description]",
-  model: "gpt-5.2-codex"  // coding specialized
-})
+gpt- 이 인증 아키텍처를 검토해줘: [코드]
+gpt- Review this REST API design
+지피티- 이 알고리즘 최적화 방법 알려줘
 ```
 
-### Recommended Models
-
-| Model | Use Case |
-|-------|----------|
-| gpt-5.2 | General purpose (default) |
-| gpt-5.2-codex | Coding specialized |
-| gpt-5.1-codex-mini | Quick responses needed |
+## Calling Gemini
 
-## When to Use Gemini (mcp__vibe-gemini)
+Use `gemini-` or `제미나이-` prefix:
 
-| Situation | Reason |
-|-----------|--------|
-| UI/UX design consultation | User experience perspective |
-| Design system advice | Component structuring |
-| Accessibility (a11y) review | Accessibility guidelines |
-| User flow design | UX optimization |
+| Situation | Example |
+|-----------|---------|
+| UI/UX consultation | `gemini- Improve UX for this login form` |
+| Design system advice | `gemini- Suggest component structure` |
+| Accessibility review | `gemini- Check a11y issues in this form` |
+| User flow design | `gemini- Optimize this checkout flow` |
 
-### Usage
+### Usage Examples
 
 ```
-mcp__vibe-gemini__gemini_chat({
-  prompt: "Suggest UX improvements for this login form: [description]"
-})
+gemini- 이 로그인 폼 UX 개선해줘
+gemini- Suggest improvements for this dashboard
+제미나이- 이 컴포넌트 구조 분석해줘
 ```
 
 ## When NOT to Use
@@ -63,7 +58,7 @@ mcp__vibe-gemini__gemini_chat({
 ```
 Claude: Draft design
     ↓
-GPT: Architecture review (mcp__vibe-gpt__gpt_analyze_architecture)
+gpt- Review this architecture: [design]
     ↓
 Claude: Incorporate feedback into final design
 ```
@@ -73,7 +68,7 @@ Claude: Incorporate feedback into final design
 ```
 Claude: Attempt problem analysis
     ↓
-If stuck, ask GPT (mcp__vibe-gpt__gpt_debug)
+If stuck: gpt- Debug this issue: [error details]
     ↓
 Claude: Apply GPT's suggestions
 ```
@@ -83,7 +78,7 @@ Claude: Apply GPT's suggestions
 ```
 Claude: Implement feature
     ↓
-Gemini: Suggest UX improvements (mcp__vibe-gemini__gemini_chat)
+gemini- Suggest UX improvements for: [component]
     ↓
 Claude: Apply UX feedback
 ```
@@ -91,7 +86,7 @@ Claude: Apply UX feedback
 ## API Key Setup
 
 ```bash
-vibe gpt <api-key>      # Enable GPT
-vibe gemini <api-key>   # Enable Gemini
-vibe status             # Check current settings
+vibe gpt login      # Enable GPT (OAuth)
+vibe gemini login   # Enable Gemini (OAuth)
+vibe status         # Check current settings
 ```

package/skills/tool-fallback.md

@@ -28,8 +28,8 @@ Tool State Machine:
 
 **Per-tool tracking:**
 ```
-vibe-gpt: failures=2, state=CLOSED
-vibe-gemini: failures=0, state=CLOSED
+GPT hook: failures=2, state=CLOSED
+Gemini hook: failures=0, state=CLOSED
 WebSearch: failures=3, state=OPEN (blocked until 14:32:00)
 context7: failures=1, state=CLOSED
 ```
@@ -39,7 +39,7 @@ context7: failures=1, state=CLOSED
 | Alternative | Method |
 |-------------|--------|
 | context7 MCP | `mcp__context7__query-docs` |
-| vibe-gpt | `mcp__vibe-gpt__gpt_chat` |
+| GPT hook | `gpt- [your question]` |
 | Cached knowledge | Use Claude's built-in knowledge |
 
 ```
@@ -51,7 +51,7 @@ If OPEN → Skip to alternative immediately
     ↓
 If CLOSED → Try context7 for library docs
     ↓
-If still fails, ask GPT
+If still fails: gpt- [question]
     ↓
 Last resort: Claude's built-in knowledge
 ```
@@ -78,14 +78,14 @@ Use Grep for content-based search
 Check git log for history
 ```
 
-## When MCP Fails
+## When LLM Hook Fails
 
 ```
-vibe-gpt fails
+gpt- [question] fails
     ↓
 Check circuit: If OPEN, skip to next
     ↓
-Try vibe-gemini (similar capability)
+Try gemini- [question] (similar capability)
     ↓
 Try context7 (for docs)
     ↓

package/skills/vibe-capabilities.md

@@ -68,13 +68,15 @@ node -e "import('@su-record/vibe/orchestrator').then(o =>
 )"
 ```
 
-## MCP Integrations
+## LLM Integrations (Hooks)
 
-| MCP | Purpose |
-|-----|---------|
-| vibe-gpt | GPT sub-agent (architecture, debugging) |
-| vibe-gemini | Gemini sub-agent (UI/UX) |
-| context7 | Latest library documentation |
+Use prefix patterns to call GPT/Gemini:
+
+| LLM | Prefix | Example |
+|-----|--------|---------|
+| GPT | `gpt-` or `지피티-` | `gpt- Review this code` |
+| Gemini | `gemini-` or `제미나이-` | `gemini- Analyze UX` |
+| context7 | MCP call | Library documentation |
 
 ## ULTRAWORK Mode
 
@@ -122,6 +124,6 @@ Add `ultrawork` or `ulw` for maximum performance:
 Even without `/vibe.*` commands, you can:
 
 1. Call tools directly via node commands
-2. Use MCP integrations (gpt, gemini, context7)
+2. Use LLM hooks (`gpt-`, `gemini-`)
 3. Reference skills for guidance
 4. Apply coding rules from `.claude/vibe/rules/`