@su-record/vibe 2.16.4 → 3.0.0

package/skills/test/SKILL.md

@@ -72,10 +72,10 @@ Written to `~/.vibe/test-reports/<YYYYMMDD-HHmm>-<harness>.{json,md}`. Exact sch
     ],
     "hooks": [
       { "name": "pre-tool-guard", "status": "pass", "tests": "38/38" },
-      { "name": "keyword-detector", "status": "pass", "tests": "12/12" }
+      { "name": "scope-guard", "status": "pass", "tests": "12/12" }
     ],
     "agents": [
-      { "name": "explorer", "status": "pass" }
+      { "name": "implementer", "status": "pass" }
     ]
   },
   "failed": [
@@ -111,7 +111,7 @@ Written to `~/.vibe/test-reports/<YYYYMMDD-HHmm>-<harness>.{json,md}`. Exact sch
 ## Failures
 
 - **skills / spec** — invocation includes `auto` but triggers array is empty
-- **agents / explorer** — agent file not found
+- **agents / implementer** — agent file not found
 ```
 
 If `failed` is empty, replace the Failures section with `_All probes passed._`.

package/skills/tool-fallback/SKILL.md

@@ -3,8 +3,8 @@ name: tool-fallback
 invocation: [auto]
 tier: optional
 description: "Tool failure fallback strategies with circuit breaker. Auto-activates on API errors, search failures, timeouts, 429, 5xx, overloaded errors."
-triggers: [API error, search failure, timeout, 429, 5xx, overloaded, fallback, circuit breaker]
-priority: 80
+triggers: [search failure, 429, 5xx, overloaded, fallback, circuit breaker, rate limit]
+priority: 60
 ---
 
 # Tool Fallback Strategies

package/skills/vercel-react-best-practices/SKILL.md

@@ -3,7 +3,7 @@ name: vercel-react-best-practices
 invocation: [auto]
 tier: core
 description: "React/Next.js performance gotchas from Vercel engineering. Non-intuitive pitfalls that LLMs commonly miss."
-triggers: [react, next.js, performance, optimization, vercel, component, rendering]
+triggers: [react, next.js, vercel, react performance, re-render, rendering waterfall]
 priority: 60
 ---
 

package/skills/vibe/SKILL.md

@@ -31,7 +31,7 @@ user-invocable: true
 - **동적 파이프라인**: 의도/입력 종류에 따라 매번 다른 스킬 체인 구성. 미리 정해진 고정 흐름 아님.
 - **무제한 라우팅**: 라우팅 표는 빠른 경로일 뿐 닫힌 화이트리스트가 아니다. 설치된 모든 `vibe.*` 스킬이 라우팅 후보이며, 표에 없는 요구사항도 description 기반 의미 매칭으로 처리한다 (Catch-all).
 - **하네스 정규화 (추론 앞단)**: vibe는 CC(추론)·Codex(직역) 어느 하네스의 암묵적 동작에도 의존하지 않는다. `/vibe`가 모호한 NL을 **명시적·직역 가능한 지시로 먼저 전개**하고, 하위 skill은 모호한 입력을 받지 않는다. 이로써 모든 하네스에서 동일 결과 + CC급 편의를 제공한다. 전문: `vibe/rules/principles/dual-harness-doctrine.md`.
-- **Smart Resume**: `.vibe/{interviews,plans,specs,features}/` 감지하여 "이어서 진행?" 자동 제안.
+- **Smart Resume**: `.vibe/{specs,features}/` 감지하여 "이어서 진행?" 자동 제안 (레거시 `.vibe/{interviews,plans}/` 는 입력 컨텍스트로만 인식).
 - **SPEC 확정 1회 승인 → 루프**: SPEC(Done의 정의 + 수용 기준)을 사용자와 확정하는 것이 유일한 의무적 사람 개입. 승인 후에는 `vibe/rules/loop-contract.md`의 ANCHOR→ACT→JUDGE→RECORD 루프를 게이트 전부 통과할 때까지 자동 반복한다. 루프 종료 = 게이트 통과 또는 stuck 또는 max_iterations 도달.
 - **위임자 역할**: `/vibe` 본인은 코드를 직접 쓰지 않는다. 라우팅·설계·실행 위임만 한다.
 
@@ -56,8 +56,8 @@ user-invocable: true
 
 | 의도 | 신호 | 기본 파이프라인 |
 |---|---|---|
-| **new feature** | "만들어줘", "추가", 신규 아이디어, 입력 없이 빈 호출 | interview → plan → spec → review → run → verify → contract → trace |
-| **figma-driven UI** | figma.com URL, "디자인", 이미지/PDF + UI 단어 | figma-extract → figma-convert → run → verify |
+| **new feature** | "만들어줘", "추가", 신규 아이디어, 입력 없이 빈 호출 | spec → run → verify (trace/contract 는 사용자 요청 또는 SPEC 에 API Contract 섹션이 있을 때만) |
+| **figma-driven UI** | figma.com URL, "디자인", 이미지/PDF + UI 단어 | figma (Extract → Convert Mode) → run → verify |
 | **clone existing UI** | 일반 웹사이트 URL + "비슷한", "클론", "이런 느낌" | clone → run → verify |
 | **resume in-progress** | feature name + `.vibe/specs/{name}.md` 존재 | (resume 지점부터) |
 | **review only** | "리뷰", "검토만", 코드/SPEC 첨부 + "확인" | review |
@@ -92,18 +92,17 @@ user-invocable: true
 .vibe/.last-feature 존재 → 직전 feature 이름 추출
 .vibe/specs/{feature}/ 또는 .vibe/specs/{feature}.md 존재 → spec 단계 완료
 .vibe/features/{feature}/ 또는 .vibe/features/{feature}.feature 존재 → feature 단계 완료
-.vibe/plans/{feature}.md 존재 → plan 단계 완료
-.vibe/interviews/{feature}.md 존재 → interview 단계 완료
+.vibe/plans/{feature}.md · .vibe/interviews/{feature}.md 존재 → 레거시 아티팩트 (구버전) — spec 패스의 입력 컨텍스트로만 사용, 재생성 금지
 ```
 
 감지된 진행 상태가 있으면 사용자에게 명시:
 
 ```
 🔍 진행 중인 작업 감지: "login"
-   ✅ interview · ✅ plan · ✅ spec · ⏳ review
+   ✅ spec · ⏳ run
 
-이어서 review 부터 진행할까요? 아니면 새 작업?
-[1] 이어서 review
+이어서 run 부터 진행할까요? 아니면 새 작업?
+[1] 이어서 run
 [2] 새 작업 (위 요구사항으로)
 [3] 취소
 ```
@@ -126,32 +125,21 @@ Input: 자연어 ("패럴랙스 웹사이트 만들어줘")
 Resume: 없음 (신규)
 Keywords: 없음
 
-Phase 1: /vibe.spec → interview → plan → spec → review
+Phase 1: /vibe.spec → 단일 패스 SPEC + 승인 (유일한 의무 게이트)
 Phase 2: /vibe.figma → UI 디자인 트랙 (type=website 감지)
 Phase 3: /vibe.run  → 구현
 Phase 4: /vibe.verify → 검증
-Phase 5: /vibe.contract → API 계약 검증
-Phase 6: /vibe.trace → RTM 출력
 
-예상: 6 phase, 중단 게이트 2회 (interview 후, plan 후)
+예상: 4 phase, 사람 개입 1회 (SPEC 승인)
 ```
 
-### Phase 4: 승인 게이트
+설계된 파이프라인은 요약 출력 후 바로 실행한다 — **별도의 파이프라인 승인 게이트는 없다.** 유일한 의무적 사람 개입은 spec 단계 안의 SPEC 확정 승인이다 (loop-contract SSOT).
 
-```
-위 파이프라인으로 진행할까요?
-[1] 진행
-[2] 수정 (어느 phase 빼거나 추가)
-[3] 취소
-```
-
-**Skip 조건:**
-- `automationLevel: autonomous` 설정 (`.vibe/config.json`) 또는 `ultrawork`/`ulw` 별칭 → SPEC 승인 게이트만 skip, 루프는 정상 동작
+**SPEC 승인 skip 조건:**
+- `automationLevel: autonomous` 설정 (`.vibe/config.json`) 또는 `ultrawork`/`ulw` 별칭 → SPEC 승인 게이트 skip, 루프는 정상 동작
 - `--max-iter 1` 또는 `quick` 별칭 → 1회 시도 후 종료
 
-> 참고: Phase 4의 승인 게이트는 **파이프라인 설계 승인**이다. SPEC 확정 시점의 승인(loop-contract의 "유일한 의무 개입")과 구별된다.
-
-### Phase 5: 체인 실행
+### Phase 4: 체인 실행
 
 승인된 파이프라인을 순차 호출:
 
@@ -166,18 +154,18 @@ Phase 6: /vibe.trace → RTM 출력
 ```
 
 각 phase 종료 후 JUDGE 단계:
-- 게이트 통과 (P1=0 ∧ verifyPassed) → 루프 종료, Phase 6 보고
+- 게이트 통과 (P1=0 ∧ verifyPassed) → 루프 종료, Phase 5 보고
 - 게이트 미통과 → RECORD(run-ledger + loop-history.jsonl) 후 다음 ANCHOR로
 - stuck(연속 2회 동일 findings 해시) → `automationLevel: confirm`이면 사용자 질문; `autonomous`이면 TODO 기록 후 루프 종료
 - max_iterations(기본 10) 도달 → 잔여를 인박스로 이월
 
-### Phase 6: 종료 보고
+### Phase 5: 종료 보고
 
 ```
 ✅ /vibe 완료
 
 실행 파이프라인:
-  ✅ /vibe.spec   (interview + plan + spec + review)
+  ✅ /vibe.spec   (단일 패스 SPEC + 승인)
   ✅ /vibe.figma  (UI 디자인 트랙)
   ✅ /vibe.run    (구현)
   ✅ /vibe.verify (검증 — 9/9 통과)
@@ -203,7 +191,7 @@ Phase 6: /vibe.trace → RTM 출력
 → Intent: figma-driven UI
 → Resume: 없음
 → Pipeline:
-   1. /vibe.figma  (figma-extract + figma-convert)
+   1. /vibe.figma  (figma Extract + Convert Mode)
    2. /vibe.spec   (생성된 SPEC 자동 보정)
    3. /vibe.run
    4. /vibe.verify
@@ -290,13 +278,10 @@ Phase 6: /vibe.trace → RTM 출력
 [Phase 3: 파이프라인 설계]
 ...
 
-[Phase 4: 승인 게이트]
-사용자 응답 대기...
-
-[Phase 5: 실행]
-...
+[Phase 4: 실행]
+... (SPEC 승인은 spec 단계 내부의 1회 게이트)
 
-[Phase 6: 종료 보고]
+[Phase 5: 종료 보고]
 ...
 ```
 

package/skills/vibe.analyze/SKILL.md

@@ -87,31 +87,31 @@ Read `CLAUDE.md`, `package.json`, `pyproject.toml`, etc. to identify tech stack:
 
 #### 3. Explore Related Code (Parallel Sub-Agents)
 
-**MANDATORY: Always use explorer sub-agents. Never explore in main session.**
+**MANDATORY: Always use native Explore sub-agents. Never explore in main session.**
 
-> Why: 3 explorer-low agents return ~600 tokens of summaries to main session.
+> Why: 3 Explore agents return ~600 tokens of summaries to main session.
 > Direct Glob/Grep/Read in main session would add 5-15K tokens of raw content.
 
 **Parallel exploration (ALL in ONE message):**
 
 ```text
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Find all [FEATURE] related API endpoints. List file paths, HTTP methods, routes, and auth requirements.")
 
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Find all [FEATURE] related services, business logic, and utility functions. Map dependencies.")
 
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Find all [FEATURE] related data models, schemas, and database queries. Document relationships and key fields.")
 ```
 
 **Scale for large projects (6+ related files):**
 
 ```text
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Find all test files related to [FEATURE]. Identify tested vs untested paths.")
 
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Analyze [FEATURE] configuration, environment variables, and external integrations.")
 ```
 
@@ -430,8 +430,6 @@ node -e "import('{{VIBE_PATH_URL}}/node_modules/@su-record/vibe/dist/tools/index
 
 | Tool | Purpose | When to Use |
 |------|---------|-------------|
-| `findSymbol` | Find symbol definitions | Locate function/class implementations |
-| `findReferences` | Find all references | Track usage patterns |
 | `analyzeComplexity` | Complexity analysis | Measure complexity metrics |
 | `validateCodeQuality` | Quality validation | Check code quality standards |
 | `saveMemory` | Save analysis results | Persist findings for future sessions |

package/skills/vibe.docs/SKILL.md

@@ -16,7 +16,7 @@ Generate or update project documentation by analyzing the actual codebase.
 /vibe.docs guide      # 사용자 가이드 (docs/GUIDE.md)
 /vibe.docs arch       # 아키텍처 문서 + Mermaid 다이어그램 (docs/ARCHITECTURE.md)
 /vibe.docs release    # 릴리즈 노트 (git history 기반)
-/vibe.docs agent      # CLAUDE.md / AGENTS.md 생성·동기화 (claude-md-guide + agents-md 검증 필수)
+/vibe.docs agent      # CLAUDE.md / AGENTS.md 생성·동기화 (agents-md 검증 필수)
 ```
 
 ## Pipeline Position

package/skills/vibe.event/SKILL.md

@@ -42,11 +42,11 @@ Automate community event planning with D-Day based proactive execution.
    └── DWK: 9 steps (D-30 → D+3)
 
 5. AUTO-EXECUTE pending deliverables
-   ├── Content drafts → event-content agent
-   ├── Image generation → event-image agent
-   ├── SMS/email drafts → event-comms agent
-   ├── Ops materials → event-ops agent
-   └── Speaker research → event-speaker agent
+   ├── Content drafts → event-planner agent
+   ├── Speaker research → event-planner agent
+   ├── Image generation → event-ops agent
+   ├── SMS/email drafts → event-ops agent
+   └── Ops materials → event-ops agent
 
 6. REQUEST confirmation for send actions
    ├── SMS actual send
@@ -73,41 +73,41 @@ Automate community event planning with D-Day based proactive execution.
 
 | Phase | Agents Triggered | Outputs |
 |-------|-----------------|---------|
-| D-60 | event-speaker | 3 topic suggestions + speaker research |
-| D-45 | event-speaker | DM + email outreach drafts |
-| D-40 | event-comms | Topic confirmation request email |
-| D-28 | event-content, event-image, event-comms | Intro + 2 images + 3 SMS + 1 email |
-| D-30 | event-content, event-comms | LinkedIn/Threads promo + SMS send(✅) |
-| D-14 | event-content | LinkedIn/Threads reminder |
-| D-3 | event-comms | BCC email + SMS send(✅) |
+| D-60 | event-planner | 3 topic suggestions + speaker research |
+| D-45 | event-planner | DM + email outreach drafts |
+| D-40 | event-ops | Topic confirmation request email |
+| D-28 | event-planner, event-ops | Intro + 2 images + 3 SMS + 1 email |
+| D-30 | event-planner, event-ops | LinkedIn/Threads promo + SMS send(✅) |
+| D-14 | event-planner | LinkedIn/Threads reminder |
+| D-3 | event-ops | BCC email + SMS send(✅) |
 | D-1 | event-ops | Nametags + checklist + PPTX + SMS(✅) |
-| D-Day | event-comms | Day-of SMS(✅) |
-| D+1 | event-content, event-ops | Review posts + settlement |
+| D-Day | event-ops | Day-of SMS(✅) |
+| D+1 | event-planner, event-ops | Review posts + settlement |
 | D+2 | — | Site update (manual) |
 
 ### Webinar Phases
 
 | Phase | Agents Triggered | Outputs |
 |-------|-----------------|---------|
-| Planning | event-content, event-image, event-comms | Intro + 3 images + SMS + email |
-| Planning | event-content | LinkedIn/Threads promo |
-| D-4 | event-comms | Zoom(manual) + BCC email + SMS(✅) |
-| D-Day | event-comms, event-ops | SMS(✅) + slides |
-| D+1 | event-content | Review posts |
+| Planning | event-planner, event-ops | Intro + 3 images + SMS + email |
+| Planning | event-planner | LinkedIn/Threads promo |
+| D-4 | event-ops | Zoom(manual) + BCC email + SMS(✅) |
+| D-Day | event-ops | SMS(✅) + slides |
+| D+1 | event-planner | Review posts |
 
 ### DWK Phases
 
 | Phase | Agents Triggered | Outputs |
 |-------|-----------------|---------|
-| D-30 | event-speaker | 3 topics + speaker research |
-| D-25 | event-speaker | DM + email outreach |
-| D-20 | event-comms | Topic confirmation email |
-| D-16 | event-content, event-image, event-comms | Intro + 2 images + SMS + email |
-| D-15 | event-content | LinkedIn/Threads promo |
-| D-3 | event-comms | BCC email + SMS(✅) |
+| D-30 | event-planner | 3 topics + speaker research |
+| D-25 | event-planner | DM + email outreach |
+| D-20 | event-ops | Topic confirmation email |
+| D-16 | event-planner, event-ops | Intro + 2 images + SMS + email |
+| D-15 | event-planner | LinkedIn/Threads promo |
+| D-3 | event-ops | BCC email + SMS(✅) |
 | D-1 | event-ops | Slido plan + PPTX + SNS remind(✅) |
-| D-Day | event-comms, event-ops | SMS(✅) + nametags(offline) |
-| D+3 | event-content, event-ops | Review + settlement |
+| D-Day | event-ops | SMS(✅) + nametags(offline) |
+| D+3 | event-planner, event-ops | Review + settlement |
 
 ## State File Format
 

package/skills/vibe.figma/SKILL.md

@@ -158,8 +158,7 @@ Load skill `figma` — Phase 1 Storyboard
 ### Phase 2 — 재료 확보
 
 ```
-Load skill `figma-extract`
-Load skill `figma` — Phase 2 (코디네이터: MO/PC 병렬 워커)
+Load skill `figma` — Phase 2 Extract Mode (코디네이터: MO/PC 병렬 워커)
 
    Phase 0에서 수집한 URL 중 "design"으로 자동 분류된 항목 (MO/PC)을 사용.
    → tree.json + bg/ + content/ + sections/ (검증용)
@@ -181,8 +180,7 @@ Load skill `figma` — Phase 3
 ### Phase 4 — BP별 스태틱 구현 (컨벤션 준수)
 
 ```
-Load skill `figma-convert`
-Load skill `figma` — Phase 4
+Load skill `figma` — Phase 4 Convert Mode
 
    재사용 매칭 (BLOCKING):
      ① component-index.json 매칭 → import (새로 만들지 않음)
@@ -191,7 +189,7 @@ Load skill `figma` — Phase 4
    ⛔ figma-to-scss.js 강제 / figma-validate.js PASS 없이 다음 섹션 금지
 
 ⤵ Phase 4 후처리
-Load skill `design-normalize`
+Load skill `design-refine` — normalize 모드
    → 매칭 안 된 하드코딩 값 → MASTER.md 토큰으로 정렬
    → 새 토큰 필요 항목은 review 큐에 기록
 ```
@@ -216,17 +214,17 @@ Load skill `figma` — Phase 6 (시각 검증 루프, P1=0까지)
      - Figma MCP/변환기가 삼킨 값이 여기서 드러난다.
 
 ⤵ Phase 6 후처리
-Load skill `design-audit`
+Load skill `design-review` — audit 모드
    → 5-dimension 점검 (a11y, performance, responsive, theming, AI slop)
    → P1 finding은 review 큐에 기록 (read-only)
 
 ⤵ Phase 6.5 (hasEmitDesignMd == true 시)
-   → resolved tokens (Color/Typography/Spacing) + design-audit 산출물을
+   → resolved tokens (Color/Typography/Spacing) + design-review(audit) 산출물을
      DESIGN.md (Stitch 9 섹션) 으로 정제해 프로젝트 루트에 저장
    → 첫 줄: `<!-- design-md-version: 1 -->`
    → `vibe.design lint` 자동 통과 확인
 
-⛔ Phase 6 + design-audit 완료 전까지 "완료 요약" 출력 금지.
+⛔ Phase 6 + design-review(audit) 완료 전까지 "완료 요약" 출력 금지.
 ```
 
 ---
@@ -260,8 +258,7 @@ Load skill `figma` — Phase 1
 ### Phase 2 — 재료 확보
 
 ```
-Load skill `figma-extract`
-Load skill `figma` — Phase 2
+Load skill `figma` — Phase 2 Extract Mode
    → Branch 1 Phase 2와 동일 (design URL로 추출)
 
    🚦 Audit gate: Branch 1과 동일하게 `auditSummary.p1 > 0` 이면 Phase 3 진입 금지.
@@ -276,14 +273,13 @@ Load skill `figma` — Phase 3 (figma-refine.js 강제)
 ### Phase 4 — BP별 스태틱 구현 (Standalone 모드)
 
 ```
-Load skill `figma-convert`
-Load skill `figma` — Phase 4
+Load skill `figma` — Phase 4 Convert Mode
 
    ⛔ 재사용 매칭 SKIP — 새 컴포넌트 자유 생성
    ✅ 새 토큰 자유 생성 (피처 스코프 네이밍: $feature-color-xxx)
    ✅ figma-to-scss.js / figma-validate.js는 그대로 강제
 
-   ❌ design-normalize 호출 금지 (이 Branch는 토큰 alignment 대상이 아님)
+   ❌ design-refine(normalize) 호출 금지 (이 Branch는 토큰 alignment 대상이 아님)
 ```
 
 ### Phase 5 — 컴파일 게이트
@@ -300,15 +296,15 @@ Load skill `figma` — Phase 6 (시각 검증 루프, P1=0까지)
    Raw-vs-computed 재조정: Branch 1과 동일 — tree.json `raw` ↔ getComputedStyle 비교.
 
 ⤵ Phase 6 후처리
-Load skill `design-audit`
+Load skill `design-review` — audit 모드
    → a11y는 항상 검증. 컨벤션 무시 모드여도 접근성은 양보 불가.
 
 ⤵ Phase 6.5 (hasEmitDesignMd == true 시)
-   → resolved tokens + design-audit 산출물을
+   → resolved tokens + design-review(audit) 산출물을
      DESIGN.md (Stitch 9 섹션) 으로 정제해 프로젝트 루트에 저장
    → 첫 줄: `<!-- design-md-version: 1 -->`
 
-⛔ Phase 6 + design-audit 완료 전까지 "완료 요약" 출력 금지.
+⛔ Phase 6 + design-review(audit) 완료 전까지 "완료 요약" 출력 금지.
 ```
 
 ---
@@ -612,7 +608,7 @@ Load skill `figma:figma-generate-design` — Step 3+: Assemble Sections (디자
 
 ```
 ❌ Branch 1 도중 figma:figma-use / figma:figma-generate-design 호출 금지 (READ 모드)
-❌ Branch 2 도중 design-normalize 호출 금지 (독립 모드)
-❌ Branch 3 도중 figma-extract / figma-convert / figma-refine.js 호출 금지 (WRITE 모드)
+❌ Branch 2 도중 design-refine(normalize) 호출 금지 (독립 모드)
+❌ Branch 3 도중 figma 스킬의 Extract/Convert Mode · figma-refine.js 호출 금지 (WRITE 모드)
 ❌ Branch 결정 후 다른 Branch의 Phase로 점프 금지
 ```

package/skills/vibe.harness/SKILL.md

@@ -30,16 +30,16 @@ When items are N/A, **remove their points from the total** rather than scoring 0
 
 ### 1. Collect Project State (Parallel Agents)
 
-Dispatch 3 explorer agents in a single message:
+Dispatch 3 native Explore agents in a single message:
 
 ```text
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Check project scaffolding: 1) Does docs/ exist with business documents? 2) Does .dev/ exist for AI logs? 3) Is src/ organized by role (not flat)? 4) Is tests/ separate from src/? 5) List top-level directory structure.")
 
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Check project context and boundaries: 1) Does CLAUDE.md exist? How many lines? 2) Does .claude/rules/ or .vibe/ exist? How many rule files? 3) Does .claude/settings.local.json exist with hooks? 4) Does .vibe/config.json exist? 5) Are there any .claude/skills/ directories?")
 
-Agent(subagent_type="explorer-low", model="haiku",
+Agent(subagent_type="Explore", model="haiku",
   prompt="Check project planning, execution, and verification: 1) Are there SPEC files in .vibe/specs/? 2) Are there Feature (BDD) files in .vibe/features/? 3) Are there test files? How many? 4) Is there CI config (.github/workflows, etc.)? 5) Are there .dev/learnings/ files?")
 ```
 
@@ -81,7 +81,7 @@ Agent(subagent_type="explorer-low", model="haiku",
 | Item | Criteria | Points |
 |------|----------|--------|
 | Agents or skills | Specialized agents or skills defined | /5 |
-| Team composition | Agent teams (architect + implementer + tester, etc.) | /4 |
+| Parallel composition | Parallel native subagents (architect + implementer + tester, etc.) | /4 |
 | Permission model | Per-agent permission separation (read-only vs write) | /3 |
 | Non-code workflows | Support for documentation, research, and other non-code tasks | /3 |
 

package/skills/vibe.reason/SKILL.md

@@ -198,8 +198,6 @@ node -e "import('{{VIBE_PATH_URL}}/node_modules/@su-record/vibe/dist/tools/index
 
 | Tool | Purpose | When to Use |
 |------|---------|-------------|
-| `findSymbol` | Find symbol definitions | Locate potential problem areas |
-| `findReferences` | Find all references | Trace data/control flow |
 | `analyzeComplexity` | Complexity analysis | Identify high-risk complex code |
 | `saveMemory` | Save reasoning progress | Store hypothesis verification results |
 | `recallMemory` | Recall saved memory | Retrieve previous reasoning context |

package/skills/vibe.review/SKILL.md

@@ -222,78 +222,76 @@ Read CLAUDE.md         -> Explicit tech stack declaration
 5. If no SPEC files exist → Skip (reviews work without SPEC too)
 ```
 
-### Phase 2: Parallel Agent Review (STACK-AWARE) via Orchestrator
+### Phase 2: Parallel Agent Review (STACK-AWARE)
 
-**Execution via Orchestrator (12+ agents in parallel):**
-```bash
-node -e "import('{{VIBE_PATH_URL}}/node_modules/@su-record/vibe/dist/infra/orchestrator/index.js').then(o => o.review(['FILE_PATHS'], ['DETECTED_STACKS']).then(r => console.log(r.content[0].text)))"
-```
+**Spawn the reviewers as concurrent native subagents in ONE message** — one `code-reviewer` instance per focus plus `security-reviewer`, each scoped to the changed files:
 
-**Example:**
-```bash
-# Review changed files with TypeScript + React stack
-node -e "import('{{VIBE_PATH_URL}}/node_modules/@su-record/vibe/dist/infra/orchestrator/index.js').then(o => o.review(['src/api/users.ts', 'src/components/Login.tsx'], ['TypeScript', 'React']).then(r => console.log(r.content[0].text)))"
 ```
+Task (code-reviewer): "Review [FILES] — focus: correctness"      (concurrent)
+Task (code-reviewer): "Review [FILES] — focus: data-integrity"   (concurrent)
+Task (code-reviewer): "Review [FILES] — focus: performance"      (concurrent)
+Task (code-reviewer): "Review [FILES] — focus: architecture"     (concurrent)
+Task (security-reviewer): "Review [FILES] for vulnerabilities"   (concurrent)
+```
+
+Stack-specific focus (`idioms`) is added when the diff touches that stack's files. Collect all results, then dedupe/merge findings before Phase 3.
 
-**Core Reviewers (Always Run):**
-| Agent | Focus |
+**Core Reviewers (Always Run — parallel `code-reviewer` instances, one per focus, plus `security-reviewer`):**
+| Agent (focus) | Focus |
 |-------|-------|
 | security-reviewer | OWASP Top 10, vulnerabilities |
-| data-integrity-reviewer | Data validation, constraints |
-| performance-reviewer | N+1 queries, memory leaks |
-| architecture-reviewer | Layer violations, cycles |
-| complexity-reviewer | Cyclomatic complexity, length |
-| simplicity-reviewer | Over-abstraction, dead code |
-| git-history-reviewer | Churn files, risk patterns |
-| test-coverage-reviewer | Missing tests, edge cases |
-
-**Stack-Specific Reviewers (Conditional):**
-| Agent | Condition |
+| code-reviewer (focus: correctness) | Logic errors, edge cases |
+| code-reviewer (focus: data-integrity) | Data validation, constraints |
+| code-reviewer (focus: performance) | N+1 queries, memory leaks |
+| code-reviewer (focus: architecture) | Layer violations, cycles |
+| code-reviewer (focus: complexity) | Cyclomatic complexity, length, over-abstraction, dead code |
+| code-reviewer (focus: git-history) | Churn files, risk patterns |
+| code-reviewer (focus: test-coverage) | Missing tests, edge cases |
+
+**Stack-Specific Review (Conditional — one extra `code-reviewer` instance):**
+| Agent (focus) | Condition |
 |-------|-----------|
-| python-reviewer | .py files in diff |
-| typescript-reviewer | .ts/.tsx files OR tsconfig |
-| rails-reviewer | Gemfile has rails |
-| react-reviewer | package.json has react |
+| code-reviewer (focus: idioms) | Language/framework files in diff (.py / .ts / .tsx / Gemfile rails / package.json react) — prompt states the detected stack |
 
 ### Phase 2.5: UI/UX Review Agents (Auto-triggered)
 
 > **활성화 조건**: 변경된 파일 중 UI 파일 존재 (`.tsx`, `.jsx`, `.vue`, `.svelte`, `.html`, `.css`, `.scss`)
 > **비활성화**: `.vibe/config.json`에 `"uiUxAnalysis": false` 설정
 
-**기존 12+ 리뷰 에이전트와 병렬 실행:**
+**기존 Phase 2 리뷰 에이전트와 병렬 실행 — `design-reviewer` 인스턴스 3개 (관점별):**
 
-| Agent | Role | Output |
+| Agent (관점) | Role | Output |
 |-------|------|--------|
-| ⑥ ux-compliance-reviewer | UX 가이드라인 준수 검증 | P1/P2/P3 findings |
-| ⑦ ui-a11y-auditor | WCAG 2.1 AA 접근성 감사 | P1/P2/P3 findings |
-| ⑧ ui-antipattern-detector | UI 안티패턴 + 디자인 시스템 일관성 | P1/P2/P3 findings |
+| ⑥ design-reviewer (UX 준수) | UX 가이드라인 준수 검증 | P1/P2/P3 findings |
+| ⑦ design-reviewer (접근성) | WCAG 2.1 AA 접근성 감사 | P1/P2/P3 findings |
+| ⑧ design-reviewer (안티패턴) | UI 안티패턴 + 디자인 시스템 일관성 | P1/P2/P3 findings |
 
 **실행 방법 — 기존 Phase 2 에이전트와 병렬 실행:**
 
 ```text
 # ⑥ UX 준수 검증 (Haiku)
-Task(subagent_type="ux-compliance-reviewer",
+Task(subagent_type="design-reviewer",
   prompt="Review UI files for UX guideline compliance: {changed_ui_files}. Use core_ui_search against ux-guidelines and web-interface domains.")
 
 # ⑦ 접근성 감사 (Haiku)
-Task(subagent_type="ui-a11y-auditor",
+Task(subagent_type="design-reviewer",
   prompt="Audit UI files for WCAG 2.1 AA compliance: {changed_ui_files}.")
 
 # ⑧ 안티패턴 검출 (Haiku)
-Task(subagent_type="ui-antipattern-detector",
+Task(subagent_type="design-reviewer",
   prompt="Detect UI anti-patterns in: {changed_ui_files}. Check against MASTER.md if exists at .vibe/design-system/{project}/MASTER.md.")
 ```
 
 #### Visual P1 Baseline
 
 - 프로젝트 루트에 `DESIGN.md` 가 존재하면 **시각 P1 의 1 차 baseline** 으로 사용한다 (§2 Color Palette / §7 Do's & Don'ts).
-- `DESIGN.md` 부재 시 기존 폴백을 사용 (WCAG 2.1 AA + `MASTER.md` + design-audit 기본 5 차원).
+- `DESIGN.md` 부재 시 기존 폴백을 사용 (WCAG 2.1 AA + `MASTER.md` + design-review(audit 모드) 기본 5 차원).
 - v1 범위: hex 컬러 드리프트만 P1 후보. spacing / font 드리프트는 Phase 2+ 에서 추가.
 - 안티패턴 검출(⑧) 은 `DESIGN.md §7` 의 "DON'T" 항목을 우선 규칙으로 사용한다.
 
 **findings 통합**: ⑥⑦⑧ findings를 기존 findings[]와 병합 → P1/P2/P3 통합 정렬
 
-**⑦ Critical finding 에스컬레이션**: ui-a11y-auditor의 P1 finding은 Review Debate Team(Phase 4.5)에 자동 포함
+**⑦ Critical finding 에스컬레이션**: design-reviewer(접근성)의 P1 finding은 Review Debate(Phase 4.5)에 자동 포함
 
 ### Phase 2.7: Boundary Mismatch Detection (Integration Coherence)
 
@@ -362,36 +360,36 @@ P3 NICE-TO-HAVE (Enhancement) - N issues
 4. [STYLE] Consider extracting helper function
 ```
 
-### Phase 4.5: Agent Teams — Review Debate
+### Phase 4.5: Review Debate (parallel native subagents)
 
-> **팀 정의**: `agents/teams/review-debate-team.md` 참조 (Code Review 컨텍스트)
-> 설정: `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` + `teammateMode: in-process` (`~/.claude/settings.json` 전역 — postinstall 자동 설정)
+> P1/P2 findings 를 검증하기 위해 네이티브 서브에이전트를 병렬로 스폰한다 —
+> `security-reviewer` + `code-reviewer` 인스턴스(서로 다른 focus)가 각 finding 을 교차 검증(validate / upgrade / downgrade / remove)한다.
 
 **토론 결과 예시:**
 
 ```
 🤝 REVIEW DEBATE RESULTS
 
-Team Consensus (4 reviewers):
+Consensus (4 parallel reviewers):
 
 ✅ Validated P1 (unanimous):
   1. [SECURITY] SQL Injection — 4/4 agree critical
 
 ⬆️ Upgraded P2→P1 (debate result):
-  2. [PERF] Unbounded query — architecture-reviewer pointed out
-     cascading failure risk under load → team agreed P1
+  2. [PERF] Unbounded query — code-reviewer (focus: architecture) pointed out
+     cascading failure risk under load → agreed P1
 
 ⬇️ Downgraded P1→P2 (debate result):
-  3. [SECURITY] CSRF on read-only endpoint — simplicity-reviewer
-     noted endpoint has no side effects → team agreed P2
+  3. [SECURITY] CSRF on read-only endpoint — code-reviewer (focus: complexity)
+     noted endpoint has no side effects → agreed P2
 
 ❌ Removed (false positive):
-  4. [ARCH] "Circular dependency" — architecture-reviewer confirmed
+  4. [ARCH] "Circular dependency" — code-reviewer (focus: architecture) confirmed
      this is intentional bi-directional reference, not a cycle
 
-🆕 New findings (team discussion):
+🆕 New findings (cross-review discussion):
   5. [DATA] Race condition in concurrent updates — emerged from
-     security + performance discussion
+     security + performance cross-review
 ```
 
 ### Phase 5: Auto-Fix (P1/P2)
@@ -452,7 +450,7 @@ Examples:
 CODE REVIEW SUMMARY
 PR #123: Add user authentication
 
-Reviewers: 13 agents
+Reviewers: security-reviewer + 8 code-reviewer instances (per-focus) + 3 design-reviewer instances (UI)
 ⏱️ Started: {start_time}
 ⏱️ Completed: {getCurrentTime 결과}
 
@@ -519,8 +517,6 @@ node -e "import('{{VIBE_PATH_URL}}/node_modules/@su-record/vibe/dist/tools/index
 |------|---------|-------------|
 | `validateCodeQuality` | Code quality check | Overall code quality scan |
 | `analyzeComplexity` | Complexity metrics | Check function complexity |
-| `findSymbol` | Find definitions | Locate implementations |
-| `findReferences` | Find all usages | Track symbol usage |
 | `saveMemory` | Save findings | Store important review findings |
 
 ### Example Tool Usage in Review