claude-pro-minmax 1.3.1 → 1.4.0

package/.claude/CLAUDE.md

@@ -49,7 +49,8 @@ Pro Plan constraints. Every message = quota. Optimize for Pass@1.
 | Implementation | @builder | Haiku | None → Escalate |
 | Quality Review | @reviewer | Haiku | None → Escalate |
 
-@dplanner tools: `sequential-thinking`, `perplexity`, `context7`
+@dplanner tools: `sequential-thinking`, `perplexity`, `Read`, `Glob`, `Grep`
+Current library docs can be handled by the official Context7 Claude Code integration when installed.
 
 ## DEFAULT WORKFLOW
 - Simple (1-3 files): `/do` — batch plan+build+verify in one shot

package/.claude/agents/README.ko.md

@@ -12,7 +12,7 @@
 | 에이전트 | 모델 | 역할 | 도구 | 질문 |
 |---------|------|------|------|------|
 | `planner.md` | Sonnet 4.6 | 아키텍처 및 설계 결정 | Read, Glob, Grep (읽기 전용) | ≤3 (기본값 포함) |
-| `dplanner.md` | Sonnet 4.6 | 연구 기반 심층 계획 | sequential-thinking, perplexity, context7, Read, Glob, Grep | 무제한 |
+| `dplanner.md` | Sonnet 4.6 | 연구 기반 심층 계획 | sequential-thinking, perplexity, Read, Glob, Grep | 무제한 |
 | `builder.md` | Haiku 4.5 | 구현 (2-retry cap) | Read, Write, Edit, Bash, Glob, Grep | 없음 → 에스컬레이션 |
 | `reviewer.md` | Haiku 4.5 | 코드 검토 및 QA | Read, Glob, Grep (읽기 전용, 강제) | 없음 → 에스컬레이션 |
 
@@ -42,7 +42,7 @@
 **기능:**
 - `sequential-thinking`: 다단계 논리 검증
 - `perplexity`: 웹 연구 (블로그, 포럼, 최신 아티클)
-- `context7`: 라이브러리 문서 조회
+- 설치된 공식 Context7 Claude Code 통합을 통한 최신 라이브러리 문서 조회
 
 **출력 예산:** 최대 60줄 (코드 블록 제외). 출처 + 출처당 1줄 인사이트만.
 
@@ -154,7 +154,7 @@ flowchart TD
 | @planner와 @dplanner에 Sonnet 4.6 사용 | 아키텍처 결정은 추론 능력 필요. Pro Plan에서 Sonnet 4.6가 Opus 4.6보다 가성비 우수 |
 | @builder 2-retry cap | Quota 소진 방지. 2회 실패 → Sonnet 4.6/Opus 4.6으로 에스컬레이션 또는 @planner로 재설계 |
 | @reviewer 읽기 전용 강제 | Hook 기반 차단(`readonly-check.sh`). 검토 중 실수로 수정하는 것 방지 |
-| @dplanner에 MCP 도구 제공 | 연구가 많은 작업은 MCP 오버헤드 정당화 가능. `sequential-thinking` + `perplexity` + `context7`로 실패 없는 계획 가능 |
+| @dplanner에 연구 도구 + 공식 문서 통합 제공 | 연구가 많은 작업은 더 깊은 계획이 필요합니다. `sequential-thinking` + `perplexity`로 추론과 웹 조사를 처리하고, 최신 라이브러리 문서는 공식 Context7이 설치된 경우 사용할 수 있습니다 |
 | 에이전트별 출력 예산 | Output은 Input의 5배 비용 (API 가격). 엄격한 예산: builder 5줄, reviewer 1줄 PASS / 30줄 FAIL, dplanner 60줄, planner 1문장/작업 |
 | @builder 원자적 롤백 | `~/.claude/scripts/snapshot.sh`가 depth guard + 라벨 확인 포함 `git stash` 처리. 무관한 사용자 stash pop 방지. 실패 시 `pop` (또는 clean tree에서 `git checkout .`) → 즉시 에스컬레이션 가능한 깨끗한 상태. 실패당 2-4 메시지 절약은 추정치이며, API 비용은 0 |
 

package/.claude/agents/README.md

@@ -12,7 +12,7 @@ Contains sub-agent definitions for role-based task delegation with model optimiz
 | Agent | Model | Role | Tools | Questions |
 |-------|-------|------|-------|-----------|
 | `planner.md` | Sonnet 4.6 | Architecture & design decisions | Read, Glob, Grep (read-only) | ≤3 (with defaults) |
-| `dplanner.md` | Sonnet 4.6 | Deep planning with research | sequential-thinking, perplexity, context7, Read, Glob, Grep | Unlimited |
+| `dplanner.md` | Sonnet 4.6 | Deep planning with research | sequential-thinking, perplexity, Read, Glob, Grep | Unlimited |
 | `builder.md` | Haiku 4.5 | Implementation (2-retry cap) | Read, Write, Edit, Bash, Glob, Grep | None → Escalate |
 | `reviewer.md` | Haiku 4.5 | Code review & QA | Read, Glob, Grep (read-only, enforced) | None → Escalate |
 
@@ -42,7 +42,7 @@ Contains sub-agent definitions for role-based task delegation with model optimiz
 **Capabilities:**
 - `sequential-thinking`: Multi-step logic verification
 - `perplexity`: Web research (blogs, forums, latest articles)
-- `context7`: Library documentation lookup
+- Current library docs via the official Context7 Claude Code integration when installed
 
 **Output Budget:** Max 60 lines (code blocks excluded). Cite source + 1-line insight per source.
 
@@ -154,7 +154,7 @@ flowchart TD
 | Sonnet 4.6 for @planner and @dplanner | Architecture decisions need reasoning capability. Sonnet 4.6 balances cost/performance better than Opus 4.6 on Pro Plan |
 | @builder 2-retry cap | Prevents quota drain. Failed twice → Escalate to Sonnet 4.6/Opus 4.6 or @planner for re-design |
 | @reviewer read-only enforcement | Hook-based blocking (`readonly-check.sh`). Prevents accidental modifications during review |
-| @dplanner with MCP tools | Research-heavy tasks justify MCP overhead. `sequential-thinking` + `perplexity` + `context7` enable fail-proof planning |
+| @dplanner with research tools + official docs integration | Research-heavy tasks justify deeper planning. `sequential-thinking` + `perplexity` cover reasoning and web research, while current library docs can use official Context7 when installed |
 | Output Budget per agent | Output costs 5x Input (API pricing). Strict budgets: builder 5 lines, reviewer 1 line PASS / 30 lines FAIL, dplanner 60 lines, planner 1 sentence/task |
 | @builder atomic rollback | `~/.claude/scripts/snapshot.sh` handles `git stash` with depth guard + label check before `/do` execution. Prevents popping unrelated user stashes. Failure triggers `pop` (or `git checkout .` on clean tree) → clean state for immediate escalation. Estimated savings: 2-4 messages per failure, zero API cost |
 

package/.claude/agents/dplanner.md

@@ -1,9 +1,9 @@
 ---
 name: dplanner
-description: Deep Planner - Architecture specialist with extended thinking and research capabilities. Use for high-complexity architectural design.
+description: Deep Planner - Architecture specialist with extended thinking and research capabilities. Use for high-complexity architectural design and deep research.
 model: sonnet
 permissionMode: plan
-tools: sequential-thinking, perplexity, context7, Read, Glob, Grep
+tools: sequential-thinking, perplexity, Read, Glob, Grep
 disallowedTools: Write, Edit, Bash
 ---
 
@@ -15,7 +15,7 @@ You prioritize **Correctness** and **Completeness** over speed.
 ## Core Capabilities
 1.  **Sequential Thinking**: Use `sequential-thinking` to break down complex logic into verified steps.
 2.  **Web Research**: Use `perplexity` for comprehensive web research (blogs, forums, latest articles).
-3.  **Documentation Lookup**: Use `context7` to fetch full documentation for libraries.
+3.  **Documentation Lookup**: When current library docs are needed, use the official Context7 Claude Code integration if it is installed.
 
 ## When to Use
 - Designing large-scale refactoring.
@@ -25,7 +25,7 @@ You prioritize **Correctness** and **Completeness** over speed.
 ## Workflow
 1.  **Analyze**: Break down the user request.
 2.  **Think**: Use `sequential-thinking` to formulate a hypothesis or plan.
-3.  **Research**: Use `perplexity` for web research, `context7` for library docs.
+3.  **Research**: Use `perplexity` for web research, and use official Context7 when current library docs are required.
 4.  **Verify**: Cross-check your plan against constraints.
 5.  **Output**: Deliver a comprehensive, fail-proof plan.
 
@@ -38,7 +38,7 @@ You prioritize **Correctness** and **Completeness** over speed.
 - Validated Logic: ...
 
 ### 2. Research Findings
-- Source: [Context7/Docs]
+- Source: [Official docs via Context7 or web]
 - Key Insight: ...
 
 ### 3. Architecture Design
@@ -52,6 +52,7 @@ You prioritize **Correctness** and **Completeness** over speed.
 ## Rules
 - **DO NOT** guess. Verify everything.
 - **DO NOT** write code. That is for @builder.
+- Keep official Context7 command details out of this prompt; use installed integration behavior or official docs instead.
 - Use `sequential-thinking` for any logic deeper than 2 levels.
 - Maximum 60 lines output (code blocks excluded from count)
 - Research Findings: cite source + 1-line insight per source (no long quotes)

package/.claude/commands/README.ko.md

@@ -15,7 +15,7 @@
 | `do-sonnet.md` | Sonnet으로 실행 | context: fork, model: sonnet |
 | `do-opus.md` | Opus로 실행 | context: fork, model: opus |
 | `plan.md` | 복잡한 작업 | @planner → @builder 체인 |
-| `dplan.md` | 심층 계획 | @dplanner (Sequential Thinking + Perplexity + Context7) |
+| `dplan.md` | 심층 계획 | @dplanner (Sequential Thinking + Perplexity + 공식 Context7 설치 시) |
 | `review.md` | 코드 검토 | 읽기 전용, 카테고리 |
 | `learn.md` | 패턴 캡처 | 자동 추출 또는 명시적 지정 |
 | `session-save.md` | 상태 저장 | 비밀 정보 삭제 |
@@ -23,7 +23,7 @@
 | `load-context.md` | 컨텍스트 로드 | Read tool, 비용 경제 |
 | `compact-phase.md` | 전략적 축소 | 단계 인식 가지치기 |
 | `watch.md` | tmux 모니터링 | 메시지 비용 없음 |
-| `llms-txt.md` | LLM 최적화 문서 | URL에서 /llms.txt 조회 |
+| `llms-txt.md` | LLM 최적화 문서 | URL에서 raw /llms.txt 조회 |
 | `analyze-failures.md` | 도구 오류 분석 | 하이브리드 학습 (로그 + 분석) |
 
 ## 명령어 카테고리
@@ -64,7 +64,7 @@
 | 명령어 | 사용 시점 | 리소스 사용량 |
 |---------|-------------|--------------|
 | `/watch` | 장기 실행 프로세스 모니터링 | 로컬 tmux (API 사용 안 함) |
-| `/llms-txt` | LLM 최적화 문서 조회 | 최소 (WebFetch) |
+| `/llms-txt` | raw LLM 최적화 문서 조회 | 최소 (WebFetch) |
 
 ## 워크플로우 예시
 
@@ -102,7 +102,7 @@
 # 3. 연구 결과를 바탕으로 구현
 # (복잡도에 따라 /plan 또는 /do 사용)
 ```
-**Quota:** 높음 (Sonnet 4.6 + Sequential Thinking + Perplexity + Context7)
+**Quota:** 높음 (Sonnet 4.6 + Sequential Thinking + Perplexity + 공식 Context7 설치 시)
 
 ### 디버깅 및 학습
 ```bash
@@ -144,7 +144,7 @@
 | **사용 사례** | 간단한 작업 | 복잡한 로직 | 중요한 결정 | 다중 파일 기능 | 미지의 영역 |
 | **영향받는 파일** | 1-3 | 1-3 | 제한없음 | 5+ | 제한없음 |
 | **질문** | 불가 | 불가 | 불가 | ≤3 (기본값 포함) | 무제한 |
-| **연구 도구** | 불가 | 불가 | 불가 | 불가 | 가능 (Perplexity, Context7) |
+| **연구 도구** | 불가 | 불가 | 불가 | 불가 | 가능 (Perplexity + 공식 Context7 설치 시) |
 
 **결정 트리:**
 ```

package/.claude/commands/README.md

@@ -15,7 +15,7 @@ Contains slash command definitions for common workflows.
 | `do-sonnet.md` | Execute with Sonnet | context: fork, model: sonnet |
 | `do-opus.md` | Execute with Opus | context: fork, model: opus |
 | `plan.md` | Complex tasks | @planner → @builder chain |
-| `dplan.md` | Deep Planning | @dplanner (Sequential Thinking + Perplexity + Context7) |
+| `dplan.md` | Deep Planning | @dplanner (Sequential Thinking + Perplexity + official Context7 when installed) |
 | `review.md` | Code review | Read-only, categories |
 | `learn.md` | Capture patterns | Auto-extract or explicit |
 | `session-save.md` | Save state | Secret scrubbing |
@@ -23,7 +23,7 @@ Contains slash command definitions for common workflows.
 | `load-context.md` | Load context | Read tool, cost economy |
 | `compact-phase.md` | Strategic compact | Phase-aware pruning |
 | `watch.md` | tmux monitoring | Zero message cost |
-| `llms-txt.md` | LLM Documentation | Fetch /llms.txt from URLs |
+| `llms-txt.md` | LLM Documentation | Fetch raw /llms.txt from URLs |
 | `analyze-failures.md` | Analyze tool errors | Hybrid learning (log + analyze) |
 
 ## Command Categories
@@ -64,7 +64,7 @@ Contains slash command definitions for common workflows.
 | Command | When to Use | Resource Usage |
 |---------|-------------|----------------|
 | `/watch` | Monitor long-running processes | Local tmux (zero API usage) |
-| `/llms-txt` | Fetch LLM-optimized documentation | Minimal (WebFetch) |
+| `/llms-txt` | Fetch raw LLM-optimized documentation | Minimal (WebFetch) |
 
 ## Workflow Examples
 
@@ -102,7 +102,7 @@ Contains slash command definitions for common workflows.
 # 3. Implement based on research findings
 # (Use /plan or /do based on complexity)
 ```
-**Quota:** High (Sonnet 4.6 + Sequential Thinking + Perplexity + Context7)
+**Quota:** High (Sonnet 4.6 + Sequential Thinking + Perplexity + official Context7 when installed)
 
 ### Debugging & Learning
 ```bash
@@ -144,7 +144,7 @@ Contains slash command definitions for common workflows.
 | **Use Case** | Simple tasks | Complex logic | Critical decisions | Multi-file features | Unknown unknowns |
 | **Files Affected** | 1-3 | 1-3 | Any | 5+ | Any |
 | **Questions** | No | No | No | ≤3 (with defaults) | Unlimited |
-| **Research Tools** | No | No | No | No | Yes (Perplexity, Context7) |
+| **Research Tools** | No | No | No | No | Yes (Perplexity + official Context7 when installed) |
 
 **Decision Tree:**
 ```

package/.claude/commands/dplan.md

@@ -1,6 +1,6 @@
 ---
 name: dplan
-description: Deep planning with @dplanner. Use for extremely complex tasks requiring Sequential Thinking, Perplexity research, and Context7 documentation.
+description: Deep planning with @dplanner. Use for extremely complex tasks requiring Sequential Thinking, Perplexity research, and current official library docs lookup when Context7 is installed.
 argument-hint: [complex feature description]
 context: fork
 agent: dplanner
@@ -12,7 +12,7 @@ agent: dplanner
 
 ## Purpose
 Invokes the **Deep Planner** for high-complexity architectural design and research.
-Unlike `/plan`, this command utilizes `sequential-thinking`, `perplexity`, and `context7` for maximum depth.
+Unlike `/plan`, this command utilizes `sequential-thinking`, `perplexity`, and official Context7 for current library docs when it is installed.
 
 ## Arguments
 $ARGUMENTS
@@ -23,7 +23,7 @@ $ARGUMENTS
    - Analyzes request.
    - Uses Sequential Thinking for logic verification.
    - Uses Perplexity for web research.
-   - Uses Context7 for library documentation.
+   - Uses official Context7 when current library documentation is needed.
    - Produces a comprehensive "Deep Plan".
 3. Returns plan to main conversation.
 

package/.claude/commands/llms-txt.md

@@ -1,6 +1,6 @@
 ---
 name: llms-txt
-description: Fetch llms.txt documentation for a library. Use when Context7 doesn't cover a library.
+description: Fetch raw llms.txt documentation for a library or direct URL. Use when you explicitly want raw /llms.txt content or when official Context7 is unavailable.
 argument-hint: nextjs | prisma | supabase | [custom-url]
 allowed-tools: Bash(curl:*)
 ---
@@ -38,9 +38,9 @@ curl -s https://nextjs.org/llms.txt | head -200
 ```
 
 ## When to Use
-- Context7 MCP가 해당 라이브러리를 찾지 못할 때
-- 빠른 원샷 문법 확인이 필요할 때
-- 최신 API 문서가 필요할 때
+- You explicitly want the raw `/llms.txt` file
+- You want a direct URL-based docs fetch
+- You need a quick one-shot raw-doc pull instead of official Context7
 
 ## Examples
 ```

package/.claude/skills/README.ko.md

@@ -28,7 +28,7 @@ skills/
 
 | 스킬 | 용도 | 주요 이점 |
 | :--- | :--- | :--- |
-| `cli-wrappers` | 무거운 MCP 서버를 가벼운 CLI 패턴으로 대체 | 출력을 상당히 줄임 |
+| `cli-patterns` | 무거운 MCP 서버를 가벼운 CLI 패턴으로 대체 | 출력을 상당히 줄임 |
 | `learned` | `/learn` 명령어를 통해 학습된 패턴 저장 | 재작업 및 반복 질문 방지 |
 
 ## 커스텀 스킬 추가 방법
@@ -57,4 +57,4 @@ Claude는 아래 조건이 충족될 때만 `read_file` 도구를 사용하여
 ## 권장 사항
 1. **50줄 규칙**: `SKILL.md`가 50줄을 넘어가면 세부 내용을 `references/`로 옮기세요.
 2. **명시적 트리거**: **"Claude는 반드시 ...을 로드해야 합니다"**와 같은 강조 문구를 사용하여 모델이 패턴을 따르게 하세요.
-3. **상대 경로**: 도구 실행 실패를 방지하기 위해 항상 프로젝트 루트부터의 전체 경로(`.claude/skills/...`)를 제공하세요.
+3. **상대 경로**: 도구 실행 실패를 방지하기 위해 항상 프로젝트 루트부터의 전체 경로(`.claude/skills/...`)를 제공하세요.

package/.claude/skills/README.md

@@ -28,7 +28,7 @@ Instead of loading extensive documentation every time, Claude starts with only t
 
 | Skill | Purpose | Key Benefit |
 | :--- | :--- | :--- |
-| `cli-wrappers` | Replaces heavy MCP servers with slim CLI patterns. | Significantly reduces output tokens. |
+| `cli-patterns` | Replaces heavy MCP servers with slim CLI patterns. | Significantly reduces output tokens. |
 | `learned` | Stores project-specific patterns via `/learn`. | Prevents rework & repeat questions. |
 
 ## Adding a Custom Skill
@@ -57,4 +57,4 @@ Contains the heavy stuff: Full API docs, long code examples, error message table
 ## Best Practices
 1. **The 50-Line Rule**: If `SKILL.md` exceeds 50 lines, move details to `references/`.
 2. **Explicit Triggers**: Use bold text like **"Claude MUST load..."** to ensure the model follows the pattern.
-3. **Relative Paths**: Always provide the full path from the project root (e.g., `.claude/skills/...`) to avoid tool failures.
+3. **Relative Paths**: Always provide the full path from the project root (e.g., `.claude/skills/...`) to avoid tool failures.

package/.claude/skills/{cli-wrappers → cli-patterns}/SKILL.md

@@ -1,12 +1,13 @@
 ---
-name: cli-wrappers
-description: Use CLI tools instead of MCP to reduce cost. Provides JSON output patterns for gh, mgrep, psql, etc.
+name: cli-patterns
+description: Use token-efficient CLI patterns instead of verbose MCP output when direct shell tools are enough. Provides JSON or compact-output conventions for gh, mgrep, psql, and similar tools.
 ---
 
-# CLI Wrappers Skill
+# CLI Patterns Skill
 
 ## Purpose
-Replace MCPs with CLI. ALWAYS use JSON output + jq.
+Prefer compact, stable CLI output over verbose MCP output when direct shell tools are enough.
+Use JSON + jq when a tool supports it, and otherwise use the smallest practical text or CSV output mode.
 
 ## GitHub (gh)
 ```bash

package/.claude.json

@@ -7,13 +7,6 @@
                 "@modelcontextprotocol/server-sequential-thinking"
             ]
         },
-        "context7": {
-            "command": "npx",
-            "args": [
-                "-y",
-                "@upstash/context7-mcp"
-            ]
-        },
         "_perplexity_disabled_by_default": {
             "command": "npx",
             "args": [
@@ -25,4 +18,4 @@
             }
         }
     }
-}
+}

package/README.ko.md

@@ -16,7 +16,7 @@ CPMM은 모델 라우팅, 출력 제어, 로컬 안전장치로 리셋 전까지
 
 > **설치 완료했다면 여기서 시작하세요: [사용자 가이드](docs/USER-MANUAL.ko.md)**
 >
-> **New in v1.3.1:** RTK를 이미 켜둔 사용자에 대해 `cpmm setup`이 관리된 hook 순서와 timeout을 자동 복원합니다.
+> **New in v1.4.0:** `cpmm setup`에서 `ctx7` 전역 설치 후 공식 Context7 Claude Code 설정까지 opt-in으로 실행할 수 있고, `/llms-txt`는 explicit raw-doc fallback으로 유지됩니다.
 
 ---
 
@@ -52,22 +52,38 @@ cpmm setup
 cpmm doctor
 ```
 
-> **v1.3.1 참고:** `cpmm setup`은 지원 환경에서 RTK 설치를 계속 시도합니다. RTK 활성화는 여전히 opt-in이지만, 한 번 켠 뒤에는 CPMM이 관리된 hook 순서와 timeout을 자동 복원합니다.
+### 4. 선택: 최신 라이브러리 문서 설정
+
+이제 interactive `cpmm setup`에서 권장 Context7 경로 전체를 opt-in으로 실행할 수 있습니다.
+
+공식 수동 설정 경로:
+```bash
+npm install -g ctx7
+ctx7 setup --cli --claude
+```
+
+- CPMM은 Context7를 기본 MCP 경로로 취급하지 않습니다.
+- 공식 설정 후에는 Context7의 공식 문서 통합이 최신 라이브러리 문서 조회를 처리합니다.
+- `/llms-txt`는 raw `/llms.txt` 내용이나 URL 기반 raw-doc 조회가 필요할 때만 사용하세요.
+
+> **v1.4.0 참고:** `cpmm setup`은 지원 환경에서 RTK 설치를 계속 시도합니다. RTK 활성화는 여전히 opt-in이며, Context7 opt-in은 `ctx7` 전역 설치와 공식 Claude Code 설정을 함께 진행합니다.
 
 의존성 정책:
 - `required`: `jq`, `mgrep`, `tmux`
+- `optional` (interactive opt-in): `ctx7` + 공식 Context7 Claude 통합
 - `optional` (자동 설치 시도): `rtk`
 - `optional` (확인만): `claude` (사전 설치 가정)
 - 도구별 자동 설치 경로:
   - `mgrep`: `npm`
+  - `ctx7`: `npm` + `ctx7 setup --cli --claude` (`cpmm setup`에서 interactive opt-in)
   - `rtk`: `brew` 또는 upstream `curl` installer
   - `jq`, `tmux`: `brew` (macOS) 또는 Linux 패키지 매니저 `apt-get`, `dnf`, `pacman`, `apk`
 - macOS에서 Homebrew가 없으면 설치 명령을 안내합니다
 
-### 4. 커스텀 & 업데이트 정책
+### 5. 커스텀 & 업데이트 정책
 
-- `cpmm setup`은 누락된 의존성을 설치한 뒤, CPMM 설정(설정 파일 복사, 언어 선택, Perplexity 설정)까지 진행합니다.
-- `cpmm doctor`는 수정 없이 의존성 상태와 RTK hook 상태를 확인합니다.
+- `cpmm setup`은 누락된 의존성을 설치한 뒤, CPMM 설정(설정 파일 복사, 언어 선택, Perplexity 설정, optional `ctx7` 설치 + 공식 Context7 설정, managed config 정리)까지 진행합니다.
+- `cpmm doctor`는 수정 없이 의존성 상태, Context7 상태, RTK hook 상태를 확인합니다.
 - 재실행 시 CPMM 관리 파일은 최신 버전으로 교체되고, 사용자 데이터는 보존됩니다.
 
 ```text
@@ -76,7 +92,7 @@ cpmm doctor
   ├── commands/          🔄 업데이트 시 교체됨
   ├── contexts/          🔄 업데이트 시 교체됨
   ├── scripts/           🔄 업데이트 시 교체됨
-  ├── skills/cli-wrappers/ 🔄 업데이트 시 교체됨
+  ├── skills/cli-patterns/ 🔄 업데이트 시 교체됨
   ├── rules/*.md         🔄 업데이트 시 교체됨
   ├── settings.json      🔄 업데이트 시 교체됨
   ├── settings.local.json  ✋ 사용자 소유 — 보존됨
@@ -98,10 +114,15 @@ cpmm doctor
 > 1. 글로벌 커스텀은 일반적으로 `settings.local.json`에 둡니다. `settings.json`은 CPMM 관리 대상이라 업데이트 시 덮어쓰기되므로, RTK 같은 third-party hook을 여기에 넣었다면 업데이트 후 다시 확인해야 합니다.
 > 2. 커스텀 명령어/규칙은 프로젝트 `.claude/`에 — 글로벌 `commands/`는 CPMM이 관리합니다.
 
+관리 설정 경계:
+- `~/.claude.json`은 CPMM 관리 대상이며 업데이트 시 정리될 수 있습니다.
+- regular file `~/.mcp.json`은 사용자 소유라 CPMM이 자동 수정하지 않습니다.
+- symlink `~/.mcp.json`은 CPMM 관리 호환 경로로 취급합니다.
+
 프로젝트 초기화 팁:
 - `claude` 실행 전에 `project-templates/`를 참고해 프로젝트를 초기화하세요. (설치기는 `project-templates`를 `~/.claude`로 복사하지 않습니다.)
 
-### 5. Bash 명령 출력 필터링 (RTK)
+### 6. Bash 명령 출력 필터링 (RTK)
 
 RTK는 CPMM이 지원하는 **선택적 Bash 명령 출력 필터링 계층**입니다. `cpmm setup`은 RTK 바이너리 설치를 시도하지만, RTK hook은 기본 활성화하지 않습니다.
 
@@ -139,13 +160,20 @@ cpmm doctor
 rtk init -g --uninstall
 ```
 
-### 6. 고급 (선택)
+### 7. 고급 (선택)
 <details>
 <summary>Perplexity, 언어, 수동 설치 보기</summary>
 
-**Perplexity/언어 설정 (필수 아님):**
-- Perplexity는 `/dplan`의 웹 리서치에 사용됩니다. 설정하지 않아도 `/dplan`은 Sequential Thinking + Context7으로 동작하며, 나머지 모든 기능은 Perplexity와 무관합니다.
+**Perplexity/언어/Context7 설정 (필수 아님):**
+- Perplexity는 `/dplan`의 웹 리서치에 사용됩니다. 설정하지 않아도 `/dplan`은 Sequential Thinking으로 동작하고, 최신 라이브러리 문서는 공식 Context7이 설치된 경우 사용할 수 있습니다. 나머지 모든 기능은 Perplexity와 무관합니다.
 - 최초 인터랙티브 설치 시 `cpmm setup`이 출력 언어와 Perplexity API 키를 묻습니다.
+- `cpmm setup`은 `ctx7` 전역 설치 후 공식 Context7 Claude Code 설정까지 optional 단계로 제안할 수 있습니다.
+- 수동 검증:
+  ```bash
+  command -v ctx7
+  ctx7 --version
+  cpmm doctor
+  ```
 - 영어(기본): 파일이 필요 없습니다. `~/.claude/rules/language.md`가 있으면 삭제하세요.
 - 비영어: `~/.claude/rules/language.md`를 만들어 원하는 언어를 지정하세요.
 - Perplexity를 수동으로 설정하려면 `~/.claude.json`의 `mcpServers`에 아래를 추가하세요:
@@ -273,7 +301,7 @@ flowchart LR
 | 명령어 | 설명 | 추천 상황 |
 | :--- | :--- | :--- |
 | **🧠 심층 실행** | | |
-| `/dplan [작업]` | **Sonnet 4.6** + Perplexity, Sequential Thinking, Context7 | 라이브러리 비교, 최신 기술 조사 (심층 연구) |
+| `/dplan [작업]` | **Sonnet 4.6** + Perplexity, Sequential Thinking, 공식 Context7 설치 시 | 라이브러리 비교, 최신 기술 조사 (심층 연구) |
 | `/do-sonnet` | **Sonnet 4.6**로 직접 실행 | Haiku 4.5가 계속 실패할 때 수동 격상 |
 | `/do-opus` | **Opus 4.6**으로 직접 실행 | 매우 복잡한 문제 해결 (비용 주의) |
 | **💾 세션/컨텍스트** | | |
@@ -285,7 +313,7 @@ flowchart LR
 | `/learn` | 패턴 학습 및 저장 | 자주 반복되는 오류나 선호 스타일 등록 |
 | `/analyze-failures` | 오류 로그 분석 | 반복되는 에러 원인 파악 |
 | `/watch` | 프로세스 모니터링 (tmux) | 장시간 빌드/테스트 관찰 |
-| `/llms-txt` | 문서 가져오기 | 라이브러리 공식 문서를 LLM 포맷으로 로드 |
+| `/llms-txt` | raw 문서 가져오기 | raw `/llms.txt` 내용 또는 직접 docs URL 로드 |
 
 </details>
 
@@ -324,7 +352,7 @@ flowchart LR
 | **🕹️ Commands** | /plan, /do, /review 등 14개 명령어 사용법 | [📂 **Commands 가이드**](.claude/commands/README.ko.md) |
 | **🪝 Hooks** | Pre-check, Auto-format 등 11개 자동화 스크립트 로직 | [📂 **Hooks 가이드**](scripts/hooks/README.ko.md) |
 | **📏 Rules** | 보안, 코드 스타일, 위험 명령어 차단 정책 | [📂 **Rules 가이드**](.claude/rules/README.ko.md) |
-| **🧠 Skills** | CLI Wrapper 등 도구 기술 명세 | [📂 **Skills 가이드**](.claude/skills/README.ko.md) |
+| **🧠 Skills** | CLI Patterns 같은 도구 스킬 사양 | [📂 **Skills 가이드**](.claude/skills/README.ko.md) |
 | **🔧 Contexts** | Backend/Frontend 프로젝트별 컨텍스트 템플릿 | [📂 **Contexts 가이드**](.claude/contexts/README.ko.md) |
 | **💾 Sessions** | 세션 요약 저장 및 관리 구조 | [📂 **Sessions 가이드**](.claude/sessions/README.ko.md) |
 | **🛠️ Scripts** | Verify, Build, Test 범용 스크립트 모음 | [📂 **Scripts 가이드**](scripts/README.ko.md) |
@@ -341,7 +369,7 @@ flowchart LR
 
 ```text
 claude-pro-minmax
-├── .claude.json                # 글로벌 MCP 설정 (User Scope)
+├── .claude.json                # 관리되는 글로벌 MCP 설정
 ├── .claudeignore               # Claude 컨텍스트 제외 규칙
 ├── .gitignore                  # Git ignore 규칙
 ├── CONTRIBUTING.md             # 기여 가이드
@@ -377,13 +405,13 @@ claude-pro-minmax
 │   │   ├── load-context.md     # 사전 정의된 컨텍스트 템플릿 로드
 │   │   ├── learn.md            # 새로운 패턴을 메모리에 저장
 │   │   ├── analyze-failures.md # 도구 실패 로그 분석
-│   │   └── llms-txt.md         # LLM 최적화 문서 조회
+│   │   └── llms-txt.md         # raw /llms.txt 문서 조회
 │   ├── rules/                  # 행동 규칙
 │   │   ├── critical-actions.md # 위험 명령어 차단 (rm -rf, git push -f, etc.)
 │   │   ├── code-style.md       # 코딩 컨벤션 및 표준
 │   │   └── security.md         # 보안 모범 사례
 │   ├── skills/                 # 도구 능력
-│   │   ├── cli-wrappers/       # 경량 CLI 래퍼 (MCP 오버헤드 대체)
+│   │   ├── cli-patterns/       # 경량 일반 CLI 패턴
 │   │   │   ├── SKILL.md        # 스킬 정의 및 사용법
 │   │   │   └── references/     # CLI 참조 문서
 │   │   │       ├── github-cli.md

package/README.md

@@ -16,7 +16,7 @@ CPMM helps Pro users complete more verified tasks before reset through model rou
 
 > **Already installed? Start here: [User Guide](docs/USER-MANUAL.md)**
 >
-> **New in v1.3.1:** `cpmm setup` now restores the managed RTK hook order and timeout for users who already enabled RTK.
+> **New in v1.4.0:** `cpmm setup` can now install `ctx7` globally and then run the official Context7 Claude Code setup, while keeping `/llms-txt` as an explicit raw-doc fallback.
 
 ---
 
@@ -52,22 +52,38 @@ cpmm setup
 cpmm doctor
 ```
 
-> **v1.3.1 note:** `cpmm setup` still attempts RTK installation when supported. RTK activation remains opt-in, but once enabled CPMM now restores the managed hook order and timeout automatically.
+### 4. Optional: Current Library Docs Setup
+
+During interactive `cpmm setup`, CPMM can now offer the full recommended Context7 path as an opt-in flow.
+
+Official manual setup:
+```bash
+npm install -g ctx7
+ctx7 setup --cli --claude
+```
+
+- CPMM does not treat Context7 as a default MCP path.
+- After official setup, Context7's own docs integration handles current library doc lookups.
+- Use `/llms-txt` only when you explicitly want raw `/llms.txt` content or a URL-based raw-doc fetch.
+
+> **v1.4.0 note:** `cpmm setup` still attempts RTK installation when supported. RTK activation remains opt-in, and Context7 opt-in now installs the `ctx7` CLI plus the official Claude Code integration.
 
 Dependency policy:
 - `required`: `jq`, `mgrep`, `tmux`
+- `optional` (interactive opt-in): `ctx7` + official Context7 Claude integration
 - `optional` (auto-install attempt): `rtk`
 - `optional` (check only): `claude` (assumed pre-installed)
 - auto-install paths by tool:
   - `mgrep`: `npm`
+  - `ctx7`: `npm` + `ctx7 setup --cli --claude` (interactive opt-in via `cpmm setup`)
   - `rtk`: `brew` or upstream `curl` installer
   - `jq`, `tmux`: `brew` (macOS) or Linux package managers `apt-get`, `dnf`, `pacman`, `apk`
 - on macOS without Homebrew, setup prints the Homebrew install command
 
-### 4. Customization & Update Policy
+### 5. Customization & Update Policy
 
-- `cpmm setup` installs missing dependencies, then configures CPMM (copies config files, language selection, Perplexity setup).
-- `cpmm doctor` checks dependency status and RTK hook health without modifying anything.
+- `cpmm setup` installs missing dependencies, then configures CPMM (copies config files, language selection, Perplexity setup, optional `ctx7` install + official Context7 setup, managed config cleanup).
+- `cpmm doctor` checks dependency status, Context7 status, and RTK hook health without modifying anything.
 - Re-running `cpmm setup` replaces CPMM-managed files with the latest version while preserving user data.
 
 ```text
@@ -76,7 +92,7 @@ Dependency policy:
   ├── commands/          🔄 Replaced on update
   ├── contexts/          🔄 Replaced on update
   ├── scripts/           🔄 Replaced on update
-  ├── skills/cli-wrappers/ 🔄 Replaced on update
+  ├── skills/cli-patterns/ 🔄 Replaced on update
   ├── rules/*.md         🔄 Replaced on update
   ├── settings.json      🔄 Replaced on update
   ├── settings.local.json  ✋ User-owned — preserved
@@ -98,10 +114,15 @@ Dependency policy:
 > 1. Global customization generally goes in `settings.local.json`. `settings.json` is CPMM-managed and overwritten on update; if you opt into RTK or other third-party hooks there, re-check them after updates.
 > 2. Custom commands/rules go in project `.claude/` — global `commands/` is managed by CPMM.
 
+Managed config boundary:
+- `~/.claude.json` is CPMM-managed and may be cleaned on update.
+- Regular-file `~/.mcp.json` is user-owned and CPMM leaves it untouched.
+- Symlinked `~/.mcp.json` is treated as CPMM-managed compatibility glue.
+
 Project initialization tip:
 - Before running `claude`, initialize your project with templates in `project-templates/` (not copied into `~/.claude`).
 
-### 5. Bash Command-Output Filtering (RTK)
+### 6. Bash Command-Output Filtering (RTK)
 
 CPMM supports RTK as an **optional Bash command-output filtering layer** for Bash-heavy workflows. `cpmm setup` attempts to install the RTK binary, but CPMM does **not** enable the RTK hook by default.
 
@@ -139,13 +160,20 @@ Rollback:
 rtk init -g --uninstall
 ```
 
-### 6. Advanced (Optional)
+### 7. Advanced (Optional)
 <details>
 <summary>Perplexity, language, manual install</summary>
 
-**Perplexity/Language setup (not required):**
-- Perplexity is used for web research in `/dplan`. Without it, `/dplan` still works via Sequential Thinking + Context7. All other features are unrelated to Perplexity.
+**Perplexity/Language/Context7 setup (not required):**
+- Perplexity is used for web research in `/dplan`. Without it, `/dplan` still works via Sequential Thinking, and current library docs can use official Context7 when installed. All other features are unrelated to Perplexity.
 - On fresh interactive installs, `cpmm setup` asks for output language and Perplexity API key.
+- `cpmm setup` can also offer the recommended Context7 flow: global `ctx7` install followed by official Claude Code setup.
+- For manual verification, use:
+  ```bash
+  command -v ctx7
+  ctx7 --version
+  cpmm doctor
+  ```
 - English (default): no file needed; remove `~/.claude/rules/language.md` if it exists.
 - Non-English: create `~/.claude/rules/language.md` with your preferred language.
 - To configure Perplexity manually, add this under `mcpServers` in `~/.claude.json`:
@@ -273,7 +301,7 @@ Full command list for more sophisticated tasks or session management.
 | Command | Description | Recommended Situation |
 | :--- | :--- | :--- |
 | **🧠 Deep Execution** | | |
-| `/dplan [task]` | **Sonnet 4.6** + Perplexity, Sequential Thinking, Context7 | Library comparison, latest tech research (Deep Research) |
+| `/dplan [task]` | **Sonnet 4.6** + Perplexity, Sequential Thinking, official Context7 when installed | Library comparison, latest tech research (Deep Research) |
 | `/do-sonnet` | Execute directly with **Sonnet 4.6** | Manual escalation when Haiku 4.5 keeps failing |
 | `/do-opus` | Execute directly with **Opus 4.6** | Resolving extremely complex problems (Cost caution) |
 | **💾 Session/Context** | | |
@@ -285,7 +313,7 @@ Full command list for more sophisticated tasks or session management.
 | `/learn` | Learn and save patterns | Registering frequently recurring errors or preferred styles |
 | `/analyze-failures` | Analyze error logs | Identifying causes of recurring errors |
 | `/watch` | Process monitoring (tmux) | Observing long-running builds/tests |
-| `/llms-txt` | Fetch documentation | Loading official library docs in LLM format |
+| `/llms-txt` | Fetch raw docs | Loading raw `/llms.txt` content or a direct docs URL |
 
 </details>
 
@@ -324,7 +352,7 @@ This project provides detailed documentation for each component. Refer to the li
 | **🕹️ Commands** | Usage of 14 commands including /plan, /do, /review | [📂 **Commands Guide**](.claude/commands/README.md) |
 | **🪝 Hooks** | Logic of 11 automation scripts including Pre-check, Auto-format | [📂 **Hooks Guide**](scripts/hooks/README.md) |
 | **📏 Rules** | Policies for Security, Code Style, Critical Actions | [📂 **Rules Guide**](.claude/rules/README.md) |
-| **🧠 Skills** | Technical specifications for tools like CLI Wrappers | [📂 **Skills Guide**](.claude/skills/README.md) |
+| **🧠 Skills** | Technical specifications for tools like CLI Patterns | [📂 **Skills Guide**](.claude/skills/README.md) |
 | **🔧 Contexts** | Context templates for Backend/Frontend projects | [📂 **Contexts Guide**](.claude/contexts/README.md) |
 | **💾 Sessions** | Structure for session summary storage and management | [📂 **Sessions Guide**](.claude/sessions/README.md) |
 | **🛠️ Scripts** | Collection of general-purpose scripts for Verify, Build, Test | [📂 **Scripts Guide**](scripts/README.md) |
@@ -341,7 +369,7 @@ This project provides detailed documentation for each component. Refer to the li
 
 ```text
 claude-pro-minmax
-├── .claude.json                # Global MCP Settings (User Scope)
+├── .claude.json                # Managed global MCP settings
 ├── .claudeignore               # Files excluded from Claude's context
 ├── .gitignore                  # Git ignore rules
 ├── CONTRIBUTING.md             # Contribution guide
@@ -377,13 +405,13 @@ claude-pro-minmax
 │   │   ├── load-context.md     # Load pre-defined context templates
 │   │   ├── learn.md            # Save new patterns to memory
 │   │   ├── analyze-failures.md # Analyze tool failure logs
-│   │   └── llms-txt.md         # View LLM-optimized documentation
+│   │   └── llms-txt.md         # View raw /llms.txt documentation
 │   ├── rules/                  # Behavioral Rules
 │   │   ├── critical-actions.md # Block dangerous commands (rm -rf, git push -f, etc.)
 │   │   ├── code-style.md       # Coding conventions and standards
 │   │   └── security.md         # Security best practices
 │   ├── skills/                 # Tool Capabilities
-│   │   ├── cli-wrappers/       # Lightweight CLI wrappers (Replaces MCP overhead)
+│   │   ├── cli-patterns/       # Lightweight general CLI patterns
 │   │   │   ├── SKILL.md        # Skill definition and usage
 │   │   │   └── references/     # CLI reference documentation
 │   │   │       ├── github-cli.md

package/install.sh

@@ -52,6 +52,215 @@ if [ -f "$HOME/.claude/settings.json" ] && grep -q "rtk-rewrite.sh" "$HOME/.clau
   HAS_EXISTING_RTK_HOOK=true
 fi
 
+cleanup_managed_claude_json() {
+  local target="$1"
+  local status=0
+
+  [ -f "$target" ] || return 0
+
+  set +e
+  node - "$target" <<'NODE'
+const fs = require("node:fs");
+
+const target = process.argv[2];
+const data = JSON.parse(fs.readFileSync(target, "utf8"));
+const servers = data && typeof data === "object" ? data.mcpServers : null;
+
+if (servers && Object.prototype.hasOwnProperty.call(servers, "context7")) {
+  delete servers.context7;
+  fs.writeFileSync(target, `${JSON.stringify(data, null, 4)}\n`);
+  process.exit(10);
+}
+NODE
+  status=$?
+  set -e
+
+  if [ "$status" -eq 0 ]; then
+    return 0
+  fi
+  if [ "$status" -eq 10 ]; then
+    echo "✅ Removed legacy Context7 MCP from ~/.claude.json"
+    return 0
+  fi
+
+  return "$status"
+}
+
+user_mcp_contains_legacy_context7() {
+  local target="$1"
+
+  [ -f "$target" ] || return 1
+
+  node - "$target" <<'NODE'
+const fs = require("node:fs");
+
+const target = process.argv[2];
+const data = JSON.parse(fs.readFileSync(target, "utf8"));
+const hasLegacyContext7 = Boolean(
+  data &&
+  typeof data === "object" &&
+  data.mcpServers &&
+  Object.prototype.hasOwnProperty.call(data.mcpServers, "context7")
+);
+
+process.exit(hasLegacyContext7 ? 0 : 1);
+NODE
+}
+
+context7_cli_installed() {
+  command -v ctx7 >/dev/null 2>&1
+}
+
+context7_global_bin_path() {
+  if command -v ctx7 >/dev/null 2>&1; then
+    command -v ctx7
+    return 0
+  fi
+
+  if command -v npm >/dev/null 2>&1; then
+    NPM_PREFIX=$(npm prefix -g 2>/dev/null || true)
+    if [ -n "$NPM_PREFIX" ] && [ -x "$NPM_PREFIX/bin/ctx7" ]; then
+      printf '%s\n' "$NPM_PREFIX/bin/ctx7"
+      return 0
+    fi
+  fi
+
+  return 1
+}
+
+context7_official_skill_installed() {
+  [ -f "$HOME/.claude/skills/find-docs/SKILL.md" ]
+}
+
+context7_official_rule_installed() {
+  [ -f "$HOME/.claude/rules/context7.md" ]
+}
+
+context7_official_integration_installed() {
+  context7_official_skill_installed && context7_official_rule_installed
+}
+
+print_context7_manual_hints() {
+  echo "   Re-run cpmm setup in an interactive terminal to opt in, or run the official commands directly:"
+  echo "   npm install -g ctx7"
+  echo "   ctx7 setup --cli --claude"
+}
+
+run_context7_global_install() {
+  if ! command -v npm >/dev/null 2>&1; then
+    echo "⚠️  Skipping Context7 global install (npm not found)."
+    print_context7_manual_hints
+    return 1
+  fi
+
+  echo ""
+  echo "📚 Context7 CLI Global Install"
+  echo "  $ npm install -g ctx7"
+
+  if npm install -g ctx7; then
+    CONTEXT7_BIN=$(context7_global_bin_path || true)
+    if [ -n "${CONTEXT7_BIN:-}" ]; then
+      echo "✅ Context7 CLI installed"
+      CONTEXT7_VERSION=$("$CONTEXT7_BIN" --version 2>/dev/null | head -1 || true)
+      [ -n "$CONTEXT7_VERSION" ] && echo "   Version: $CONTEXT7_VERSION"
+      return 0
+    fi
+
+    echo "⚠️  Context7 CLI installed but is not visible on PATH in this shell."
+    echo "   A later step will use the resolved global binary path directly."
+    return 0
+  fi
+
+  echo "⚠️  Context7 global install did not complete."
+  print_context7_manual_hints
+  return 1
+}
+
+run_context7_official_setup() {
+  if ! command -v npm >/dev/null 2>&1; then
+    echo "⚠️  Skipping official Context7 setup (npm not found)."
+    print_context7_manual_hints
+    return 1
+  fi
+
+  if ! context7_global_bin_path >/dev/null 2>&1; then
+    if ! run_context7_global_install; then
+      return 1
+    fi
+  fi
+
+  CONTEXT7_BIN=$(context7_global_bin_path || true)
+  if [ -z "${CONTEXT7_BIN:-}" ]; then
+    echo "⚠️  Context7 CLI binary could not be resolved after install."
+    print_context7_manual_hints
+    return 1
+  fi
+
+  echo ""
+  echo "📚 Context7 Official CLI + Skills Setup"
+  echo "   This runs the official Context7 Claude Code setup."
+  echo "  $ $CONTEXT7_BIN setup --cli --claude"
+
+  if "$CONTEXT7_BIN" setup --cli --claude; then
+    echo "✅ Context7 official CLI + Skills setup complete"
+    return 0
+  fi
+
+  echo "⚠️  Official Context7 setup did not complete."
+  print_context7_manual_hints
+  return 1
+}
+
+offer_context7_setup() {
+  local choice="${CPMM_CONTEXT7_SETUP_CHOICE:-}"
+
+  if context7_cli_installed && context7_official_integration_installed; then
+    echo ""
+    echo "📚 Context7 Setup"
+    echo "   Context7 CLI and official Claude integration are already installed."
+    return 0
+  fi
+
+  if [ -z "$choice" ] && [ ! -t 0 ]; then
+    echo ""
+    echo "📚 Context7 Setup (Optional)"
+    echo "   Skipped in non-interactive mode."
+    print_context7_manual_hints
+    return 0
+  fi
+
+  case "$choice" in
+    official-setup|skip) ;;
+    *) choice="" ;;
+  esac
+
+  if [ -z "$choice" ]; then
+    echo ""
+    echo "📚 Context7 Setup (Optional)"
+    echo "   Install ctx7 globally and run the official Context7 Claude Code setup?"
+    echo "   1) Yes, run npm install -g ctx7 + ctx7 setup --cli --claude (Recommended)"
+    echo "   2) Skip for now (default)"
+    echo -n "   Select [1-2]: "
+    read -r CONTEXT7_CHOICE < /dev/tty || CONTEXT7_CHOICE="3"
+
+    case $CONTEXT7_CHOICE in
+      1) choice="official-setup" ;;
+      *) choice="skip" ;;
+    esac
+  fi
+
+  case "$choice" in
+    official-setup)
+      run_context7_official_setup
+      ;;
+    *)
+      echo ""
+      echo "⚠️  Skipping Context7 setup for now."
+      print_context7_manual_hints
+      ;;
+  esac
+}
+
 if [ "$IS_UPDATE" = false ] && [ -d "$HOME/.claude" ]; then
   if [ -d "$HOME/.claude.pre-cpmm" ]; then
     echo "⚠️  ~/.claude.pre-cpmm already exists, skipping backup."
@@ -131,11 +340,13 @@ done
 
 shopt -u nullglob
 
-# cli-wrappers/ and scripts/ — fully CPMM managed: rm-rf to remove stale files
-if [ -d "$SCRIPT_DIR/.claude/skills/cli-wrappers" ]; then
-  rm -rf "$HOME/.claude/skills/cli-wrappers"
-  cp -R "$SCRIPT_DIR/.claude/skills/cli-wrappers" "$HOME/.claude/skills/"
-fi
+# managed skills/ and scripts/ — fully CPMM managed: rm-rf to remove stale files
+for managed_skill in cli-patterns; do
+  if [ -d "$SCRIPT_DIR/.claude/skills/$managed_skill" ]; then
+    rm -rf "$HOME/.claude/skills/$managed_skill"
+    cp -R "$SCRIPT_DIR/.claude/skills/$managed_skill" "$HOME/.claude/skills/"
+  fi
+done
 # sessions/ dir created above; example files stay in repo only (not installed)
 if [ -d "$SCRIPT_DIR/scripts" ]; then
   rm -rf "$HOME/.claude/scripts"
@@ -153,21 +364,33 @@ if [ -f "$SCRIPT_DIR/.claude.json" ]; then
       cp "$HOME/.claude.json" "$HOME/.claude.json.bak"
     fi
     cp "$SCRIPT_DIR/.claude.json" "$HOME/.claude.json"
-    echo "✅ Installed .claude.json to ~/.claude.json (User Scope)"
+    echo "✅ Installed managed ~/.claude.json"
   elif [ ! -f "$HOME/.claude.json" ]; then
     cp "$SCRIPT_DIR/.claude.json" "$HOME/.claude.json"
     echo "✅ Restored missing ~/.claude.json from repository template"
   fi
 
-  # Fresh install: enforce symlink. Update: restore only when missing (do not overwrite existing file/link).
-  if [ "$IS_UPDATE" = false ]; then
-    if [ ! -L "$HOME/.mcp.json" ] || [ "$(readlink "$HOME/.mcp.json" 2>/dev/null)" != "$HOME/.claude.json" ]; then
+  cleanup_managed_claude_json "$HOME/.claude.json"
+
+  # ~/.claude.json is CPMM-managed. Regular-file ~/.mcp.json is user-managed.
+  if [ -L "$HOME/.mcp.json" ]; then
+    if [ "$(readlink "$HOME/.mcp.json" 2>/dev/null)" != "$HOME/.claude.json" ]; then
       ln -sf "$HOME/.claude.json" "$HOME/.mcp.json"
-      echo "✅ Ensured .mcp.json → .claude.json symlink"
+      echo "✅ Ensured managed .mcp.json → .claude.json symlink"
+    fi
+  elif [ -f "$HOME/.mcp.json" ]; then
+    echo "⚠️  Leaving regular ~/.mcp.json untouched (user-managed file)."
+    if user_mcp_contains_legacy_context7 "$HOME/.mcp.json"; then
+      echo "   Detected legacy Context7 MCP in ~/.mcp.json."
+      echo "   Remove it manually if you no longer want the legacy MCP path."
     fi
-  elif [ ! -e "$HOME/.mcp.json" ] && [ ! -L "$HOME/.mcp.json" ]; then
+  else
     ln -s "$HOME/.claude.json" "$HOME/.mcp.json"
-    echo "✅ Restored missing .mcp.json → .claude.json symlink"
+    if [ "$IS_UPDATE" = true ]; then
+      echo "✅ Restored missing .mcp.json → .claude.json symlink"
+    else
+      echo "✅ Created .mcp.json → .claude.json symlink"
+    fi
   fi
 fi
 
@@ -250,6 +473,8 @@ LANGEOF
   fi
 fi
 
+offer_context7_setup
+
 # EDGE CASE #9: guard find against missing dir (scripts/ may not exist on minimal installs)
 # Make scripts executable (Recursive)
 if [ -d "$HOME/.claude/scripts" ]; then
@@ -288,9 +513,22 @@ echo "  > /dplan Analyze complex architecture"
 echo "  > /do Implement the login page"
 echo ""
 echo "Dependency Check:"
-echo "  cpmm setup       # install missing deps (jq, mgrep, tmux) + attempt optional RTK install"
+echo "  cpmm setup       # install missing deps (jq, mgrep, tmux) + offer ctx7 install + official Context7/RTK setup"
 echo "  cpmm doctor      # check status only"
 echo ""
+echo "Context7 (Optional Official Integration):"
+if context7_official_integration_installed; then
+  echo "  Installed: official Context7 Claude integration"
+else
+  echo "  Re-run cpmm setup in an interactive terminal to opt in"
+fi
+if context7_cli_installed; then
+  echo "  CLI: ctx7 ($(ctx7 --version 2>/dev/null | head -1 || echo 'installed'))"
+else
+  echo "  Manual CLI install: npm install -g ctx7"
+fi
+echo "  Manual official setup: ctx7 setup --cli --claude"
+echo ""
 if command -v rtk >/dev/null 2>&1; then
   echo "RTK (Optional Integration):"
   echo "  Installed: rtk"

package/lib/cli.js

@@ -47,6 +47,13 @@ const DEPS = [
     systemPackage: "tmux",
     description: "Terminal multiplexer for background agents",
   },
+  {
+    key: "ctx7",
+    command: "ctx7",
+    required: false,
+    installKind: "skip",
+    description: "Official Context7 CLI",
+  },
 ];
 
 function colorize(text, colorCode) {
@@ -161,6 +168,74 @@ function printStatus(results) {
   console.log("");
 }
 
+function inspectContext7Status(homeDir) {
+  const skillPath = path.join(homeDir, ".claude", "skills", "find-docs", "SKILL.md");
+  const rulePath = path.join(homeDir, ".claude", "rules", "context7.md");
+  const binaryInstalled = commandExists("ctx7");
+
+  let version = null;
+  if (binaryInstalled) {
+    try {
+      version = execSync("ctx7 --version", { encoding: "utf8", stdio: ["ignore", "pipe", "ignore"] })
+        .trim()
+        .split("\n")[0];
+    } catch {
+      version = null;
+    }
+  }
+
+  return {
+    binaryInstalled,
+    version,
+    skillInstalled: fs.existsSync(skillPath),
+    ruleInstalled: fs.existsSync(rulePath),
+  };
+}
+
+function printContext7DoctorStatus(status) {
+  console.log("\nContext7 Integration:");
+  console.log(
+    `  ${status.binaryInstalled ? colorize("✓", "32") : colorize("○", "33")} ctx7 binary       ${
+      status.binaryInstalled ? (status.version || "installed") : "not installed (optional)"
+    }`,
+  );
+  console.log(
+    `  ${status.skillInstalled ? colorize("✓", "32") : colorize("○", "33")} Context7 skill     ${
+      status.skillInstalled ? "find-docs installed" : "not installed"
+    }`,
+  );
+  console.log(
+    `  ${status.ruleInstalled ? colorize("✓", "32") : colorize("○", "33")} Context7 rule      ${
+      status.ruleInstalled ? "context7.md installed" : "not installed"
+    }`,
+  );
+
+  const fullyInstalled = status.binaryInstalled && status.skillInstalled && status.ruleInstalled;
+  const partiallyInstalled =
+    status.binaryInstalled || status.skillInstalled || status.ruleInstalled;
+
+  if (fullyInstalled) {
+    console.log("             Context7 CLI and Claude integration are fully installed.");
+    console.log("");
+    return { failed: false };
+  }
+
+  if (partiallyInstalled) {
+    console.log("             Context7 is partially installed.");
+    console.log("             Fix: run cpmm setup and opt into Context7, or run:");
+    console.log("                  npm install -g ctx7");
+    console.log("                  ctx7 setup --cli --claude");
+    console.log("");
+    return { failed: true };
+  }
+
+  console.log("             Optional. To install:");
+  console.log("                  npm install -g ctx7");
+  console.log("                  ctx7 setup --cli --claude");
+  console.log("");
+  return { failed: false };
+}
+
 function printRtkSetupResult(result) {
   if (!result) return;
 
@@ -339,12 +414,14 @@ function runDoctor() {
 
   const results = checkDeps();
   printStatus(results);
+  const context7Status = inspectContext7Status(process.env.HOME || process.env.USERPROFILE || "");
+  const context7Doctor = printContext7DoctorStatus(context7Status);
 
   const rtkStatus = inspectRtkStatus(path.join(process.env.HOME || process.env.USERPROFILE || "", ".claude", "settings.json"));
   const rtkDoctor = printRtkDoctorStatus(rtkStatus);
 
   const missing = results.filter((r) => r.required && !r.installed);
-  if (missing.length > 0 || rtkDoctor.failed) {
+  if (missing.length > 0 || rtkDoctor.failed || context7Doctor.failed) {
     console.log(`Fix: cpmm setup`);
     return 1;
   }
@@ -357,8 +434,8 @@ function printHelp() {
   console.log(`CPMM v${pkg.version}
 
 Usage:
-  cpmm setup      Install deps + configure CPMM (language, Perplexity, optional RTK)
-  cpmm doctor     Check dependency and RTK hook status
+  cpmm setup      Install deps + configure CPMM (language, Perplexity, optional ctx7 + official Context7, optional RTK)
+  cpmm doctor     Check dependency, Context7, and RTK status
   cpmm --help     Show this help
   cpmm --version  Show version
 `);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-pro-minmax",
-  "version": "1.3.1",
+  "version": "1.4.0",
   "description": "Quota-first routing and safety layer for Claude Code Pro workflows.",
   "author": "donghoon <move-hoon@users.noreply.github.com>",
   "license": "MIT",
@@ -20,7 +20,7 @@
     ".claude/commands/",
     ".claude/contexts/",
     ".claude/rules/",
-    ".claude/skills/cli-wrappers/",
+    ".claude/skills/cli-patterns/",
     ".claudeignore",
     "scripts/",
     "README.md",