claude-pro-minmax 1.0.0

package/.claude/commands/learn.md

@@ -0,0 +1,64 @@
+---
+name: learn
+description: Extract and save patterns from session. Captures recurring coding patterns, conventions, and preferences.
+argument-hint: "[pattern]" or --show
+disable-model-invocation: true
+---
+
+# /learn Command
+
+Extract and save patterns from session.
+
+## Input
+$ARGUMENTS
+
+## Execution
+1. No arguments: Analyze current session for patterns
+2. `"[pattern]"`: Save explicit pattern
+3. `--show`: Display learned patterns list
+
+## Storage Paths
+```
+~/.claude/skills/learned/YYYY-MM-DD-[pattern-name].md
+.claude/rules/conventions.md   # CONVENTION type
+.claude/rules/preferences.md   # PREFERENCE type
+```
+
+## Categories & Actions
+| Category | Storage Location |
+|----------|------------------|
+| CONVENTION | Append to `.claude/rules/conventions.md` |
+| PATTERN | Create new file in `~/.claude/skills/learned/` |
+| PREFERENCE | Append to `.claude/rules/preferences.md` |
+
+## Pattern File Format
+```markdown
+---
+name: [pattern-name]
+description: [pattern description]
+---
+
+# Pattern: [name]
+
+## Problem
+[Description of recurring issue]
+
+## Solution
+[How to solve it]
+
+## Example
+[Code example]
+```
+
+## Auto-Learning
+Stop hook detects repeated modifications:
+- Same file modified 3+ times → suggest pattern extraction
+- Use `/learn` to manually confirm and save
+
+## Examples
+```
+/learn                                    # Analyze session
+/learn "Always use try-catch for API error handling"
+/learn "Component filenames use PascalCase"
+/learn --show                             # Show list
+```

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

@@ -0,0 +1,50 @@
+---
+name: llms-txt
+description: Fetch llms.txt documentation for a library. Use when Context7 doesn't cover a library.
+argument-hint: nextjs | prisma | supabase | [custom-url]
+allowed-tools: Bash(curl:*)
+---
+
+# /llms-txt Command
+
+Fetch LLM-optimized documentation for libraries.
+
+## Library
+$ARGUMENTS
+
+## Common URLs
+| Library | URL |
+|---------|-----|
+| Next.js | https://nextjs.org/llms.txt |
+| Prisma | https://prisma.io/llms.txt |
+| Supabase | https://supabase.com/llms.txt |
+| Vercel | https://vercel.com/llms.txt |
+| Helius | https://www.helius.dev/docs/llms.txt |
+| Tailwind | https://tailwindcss.com/llms.txt |
+
+## Execution
+
+Based on the library name:
+1. Try `https://[library].org/llms.txt`
+2. Try `https://[library].io/llms.txt`
+3. Try `https://[library].com/llms.txt`
+4. Try `https://docs.[library].io/llms.txt`
+
+For custom URLs, fetch directly.
+
+```bash
+# Example
+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
+
+## Examples
+```
+/llms-txt nextjs          # Next.js 문서
+/llms-txt prisma          # Prisma 문서
+/llms-txt https://custom.dev/llms.txt  # Custom URL
+```

package/.claude/commands/load-context.md

@@ -0,0 +1,46 @@
+---
+name: load-context
+description: Load project context files. Reads backend, frontend, or session context into conversation.
+argument-hint: [type] or --list
+disable-model-invocation: false
+---
+
+# /load-context Command
+
+Load project context using Read tool.
+
+## Input
+$ARGUMENTS
+
+## Execution
+1. `[type]`: Read `.claude/contexts/[type]-context.md` file into conversation
+2. `--list`: List available context files in `.claude/contexts/`
+
+## Context Types & Files
+| Type | File |
+|------|------|
+| backend | `.claude/contexts/backend-context.md` |
+| frontend | `.claude/contexts/frontend-context.md` |
+| session | Most recent file from `.claude/sessions/` |
+
+## Cost Economy
+Warning when more than 2 contexts loaded in a session:
+```
+⚠️ WARNING: 3 contexts loaded (backend, frontend, session)
+Consider starting a new session or using /compact-phase to reduce context.
+```
+
+## Output
+```
+📂 Context loaded: [type]
+
+Content from: .claude/contexts/[type]-context.md
+[actual file content displayed]
+```
+
+## Examples
+```
+/load-context backend           # Read backend context file
+/load-context frontend          # Read frontend context file
+/load-context --list            # List available context files
+```

package/.claude/commands/plan.md

@@ -0,0 +1,62 @@
+---
+name: plan
+description: Complex task planning with @planner to @builder chain. Use when affecting 4+ files, requiring architecture decisions, unclear requirements, or new feature implementation.
+argument-hint: [feature description] or --no-build [description]
+context: fork
+agent: planner
+---
+
+# /plan Command
+
+For complex tasks. Runs in @planner subagent context.
+
+## Feature Request
+$ARGUMENTS
+
+## Options
+- Default: Plan then implement
+- `--no-build`: Plan only, no implementation
+
+## Flow
+1. This command runs in forked @planner context
+2. @planner: Architecture analysis, task breakdown
+3. Results returned to main conversation
+4. User confirmation (no auto-proceed)
+5. @builder: Implementation (after approval)
+6. @reviewer: Post-implementation review (skipped if --no-build)
+
+## When to Use
+- Affecting 4+ files
+- Architecture decisions needed
+- Unclear requirements
+- New feature implementation
+
+## When NOT to Use
+- Simple bug fix → use `/do`
+- Single file change → use `/do`
+- Clear task → use `/do`
+
+## Advanced Usage
+For extremely complex tasks requiring deep research or logic verification (e.g., verifying race conditions), consider using **`@dplanner`** directly instead of `/plan`.
+```bash
+@dplanner Analyze potential deadlocks in transaction manager
+```
+
+## Output Format
+```
+## Architecture: [feature name]
+
+### Design Decision
+[One paragraph]
+
+### Task Breakdown
+1. [task] → @builder
+2. [task] → @builder
+
+### Files Affected
+- CREATE: path/file
+- MODIFY: path/file
+
+### Proceed?
+[y/n]
+```

package/.claude/commands/review.md

@@ -0,0 +1,55 @@
+---
+name: review
+description: Code review using @reviewer agent. Checks security, type safety, performance, and logic errors. Use after code changes for quality verification.
+argument-hint: [file or directory] or --security [target] or --all
+context: fork
+agent: reviewer
+allowed-tools: Read, Grep, Glob
+---
+
+# /review Command
+
+Code review. Runs in @reviewer subagent context (read-only).
+
+## Target
+$ARGUMENTS
+
+## Options
+- Default: Review specified file/directory
+- `--security`: Security-focused review
+- `--all`: Review all changes
+
+## Execution
+This command runs in forked @reviewer context:
+- Isolated from main conversation
+- Read-only access (enforced by agent)
+- Results summarized and returned
+
+## Categories
+- **SEC**: Security (highest priority)
+- **TYPE**: Type safety
+- **PERF**: Performance
+- **LOGIC**: Logic errors
+- **STYLE**: Coding conventions
+- **TEST**: Missing tests
+
+## Output Format (Pass)
+```
+PASS - [X files reviewed, no issues]
+```
+
+## Output Format (Fail)
+```
+FAIL - [N issues]
+
+1. [SEC] file:line - [issue] → [fix]
+2. [TYPE] file:line - [issue] → [fix]
+
+Auto-fix: [command if available]
+```
+
+## Rules
+- Max 5 issues per file
+- Must include fix suggestion
+- Must check for secrets
+- **READ-ONLY** - no file modifications (enforced by context: fork + agent: reviewer)

package/.claude/commands/session-load.md

@@ -0,0 +1,61 @@
+---
+name: session-load
+description: Restore saved session state. Loads previous work context and next actions.
+argument-hint: [name] or --list
+disable-model-invocation: true
+---
+
+# /session-load Command
+
+Restore saved session state.
+
+## Session Name
+$ARGUMENTS
+
+## Execution
+1. No arguments: Load most recent session
+2. `[name]`: Load specified session
+3. `--list`: Show available sessions
+
+## Session Directory
+```
+.claude/sessions/
+├── 2025-01-28-auth-feature.md
+├── 2025-01-27-refactor-v2.md
+└── 2025-01-26-api-update.md
+```
+
+## Load Process
+1. Read session file
+2. Set completed tasks context
+3. Display next actions
+4. Prepare relevant file references
+
+## Output (Load)
+```
+📂 Loaded: [name] ([date])
+
+State: [current state description]
+Completed: [number of completed tasks]
+Next: [next action]
+
+Ready to continue.
+```
+
+## Output (List)
+```
+📋 Available sessions:
+
+1. auth-feature (2025-01-28) - JWT auth in progress
+2. refactor-v2 (2025-01-27) - API refactoring complete
+3. api-update (2025-01-26) - Endpoint additions
+
+Load with: /session-load [name]
+```
+
+## Examples
+```
+/session-load                    # Most recent
+/session-load auth-feature       # Specific session
+/session-load --list             # Show list
+```

package/.claude/commands/session-save.md

@@ -0,0 +1,79 @@
+---
+name: session-save
+description: Save session state with automatic secret scrubbing. Use to preserve work state when pausing.
+argument-hint: [name]
+disable-model-invocation: true
+---
+
+# /session-save Command
+
+Save session state with automatic secret scrubbing.
+
+## Session Name
+$ARGUMENTS
+
+## Execution
+1. Collect session state (completed tasks, next actions, context)
+2. Scrub 15+ secret patterns via `~/.claude/scripts/scrub-secrets.js`
+3. Save to session file
+
+## File Naming Options
+```bash
+# Option 1: With session ID (recommended for uniqueness)
+.claude/sessions/${CLAUDE_SESSION_ID}-[name].md
+
+# Option 2: With date (human-readable)
+.claude/sessions/YYYY-MM-DD-[name].md
+```
+
+## Security (Auto-Scrub)
+Patterns automatically removed before saving:
+- API keys (OpenAI, Stripe, GitHub, AWS, Google)
+- Database URLs with credentials
+- JWT tokens, Bearer tokens
+- Password/secret fields
+- Private keys (RSA, SSH)
+- Environment variables with secrets
+
+## Session File Format
+```markdown
+---
+session_id: ${CLAUDE_SESSION_ID}
+date: YYYY-MM-DD HH:MM
+name: [session-name]
+project: [project-path]
+---
+
+# Session: [name]
+
+## Completed Tasks
+- [completed task list]
+
+## Current State
+[current state description]
+
+## Next Actions
+- [next action list]
+
+## Context
+- Files: [relevant files]
+- Decisions: [decisions made]
+```
+
+## Output
+```
+Session saved: .claude/sessions/[filename].md
+
+Security scan:
+- Scrubbed: [N] potential secrets
+- Clean: No secrets in final file
+
+Resume with: /session-load [name]
+```
+
+## Examples
+```
+/session-save                    # Auto-generate name with session ID
+/session-save auth-feature       # Named session
+/session-save refactor-v2        # With version
+```

package/.claude/commands/watch.md

@@ -0,0 +1,58 @@
+---
+name: watch
+description: Long-running process monitoring with tmux. Watch tests, dev server, or build process. Zero message cost.
+argument-hint: tests | dev | build | custom "[cmd]"
+disable-model-invocation: true
+allowed-tools: Bash(tmux:*)
+---
+
+# /watch Command
+
+Process monitoring with tmux.
+
+## Target
+$ARGUMENTS
+
+## Execution
+```bash
+# tests: Watch test execution
+tmux new-session -d -s claude-watch-tests "~/.claude/scripts/test.sh --watch"
+
+# dev: Watch dev server
+tmux new-session -d -s claude-watch-dev "~/.claude/scripts/build.sh --dev"
+
+# build: Watch build process
+tmux new-session -d -s claude-watch-build "~/.claude/scripts/build.sh"
+
+# custom: User-defined command
+tmux new-session -d -s claude-watch-custom "[cmd]"
+```
+
+## Preset Types
+| Type | Command | Session Name |
+|------|---------|--------------|
+| tests | `~/.claude/scripts/test.sh --watch` | claude-watch-tests |
+| dev | `~/.claude/scripts/build.sh --dev` | claude-watch-dev |
+| build | `~/.claude/scripts/build.sh` | claude-watch-build |
+| custom | User-defined | claude-watch-custom |
+
+## tmux Controls
+- `Ctrl+b d`: Detach (background)
+- `tmux attach -t [name]`: Reattach
+- `tmux ls`: List sessions
+- `tmux kill-session -t [name]`: Kill session
+
+## Cost Savings
+Before: 5 check cycles = 15 messages
+With /watch: 5 check cycles = 1 message
+
+## Output
+```
+📺 Started: claude-watch-[type]
+Command: [executed command]
+
+Controls:
+- Attach: tmux attach -t claude-watch-[type]
+- Detach: Ctrl+b d
+- Kill: tmux kill-session -t claude-watch-[type]
+```

package/.claude/contexts/README.ko.md

@@ -0,0 +1,94 @@
+> **[English Version](README.md)**
+
+# Contexts Directory
+
+## 목적
+지연 로딩(lazy loading)을 위한 **프로젝트 참조 데이터**를 포함합니다. 이 파일들은 자동으로 로드되지 **않으며**, 필요할 때 `/load-context`를 사용하여 로드해야 합니다.
+
+컨텍스트 파일은 고정된 프로젝트 정보(API 엔드포인트, DB 스키마, 환경 변수)입니다. 진행 중인 작업 상태를 추적하는 세션과 다릅니다.
+
+| | Contexts | Sessions |
+|--|----------|----------|
+| **성격** | 고정된 프로젝트 참조 데이터 | 진행 중인 작업 상태 |
+| **변경 빈도** | 프로젝트 구조 변경 시만 | 매 세션마다 |
+| **예시** | API 엔드포인트, DB 스키마, 환경 변수 | 진행 상황, 결정 사항, 다음 작업 |
+| **저장 필요** | 없음 (파일 자체가 이미 저장됨) | 있음 (매번 `/session-save`) |
+
+## 내용
+
+| 파일 | 목적 | 로드 시점 |
+|------|---------|--------------|
+| `backend-context.md` | API 엔드포인트, DB 스키마, 환경 변수 | 백엔드 개발 시 |
+| `frontend-context.md` | 컴포넌트, 상태, 라우트 | 프론트엔드 개발 시 |
+
+## 사용법
+
+```bash
+# 작업을 시작할 때 로드
+/load-context backend
+
+# 사용 가능한 컨텍스트 확인
+/load-context --list
+
+# 다른 작업으로 전환
+/load-context frontend
+```
+
+## 왜 지연 로딩인가요?
+
+1. **프롬프트 캐시 효율성:** CLAUDE.md가 불변 상태로 유지되어 캐시 적중률 극대화
+2. **비용 경제:** 필요한 것만 로드 (Pro Plan: 모든 메시지 = 쿼터)
+3. **집중된 컨텍스트:** 백엔드 작업 중 프론트엔드 스키마 로드 방지
+
+## Instruction Hierarchy (지시 우선순위)
+
+> **중요:** `/load-context`는 Read tool을 사용하므로 **tool output 레벨** — Claude의 지시 우선순위에서 가장 낮은 수준으로 로드됩니다.
+
+```
+System prompt (최고)  ← CLAUDE.md, rules/, --system-prompt
+User messages         ← 사용자 프롬프트
+Tool output (최저)    ← /load-context, @file, Read tool
+```
+
+**의미:**
+- 컨텍스트 파일은 Claude가 참고하는 **참조 데이터** — tool output 레벨로 충분
+- Claude가 **반드시 따라야 할 규칙** → `.claude/rules/`에 넣으세요 (자동 로드, system prompt 레벨)
+- 고급: `claude --system-prompt "$(cat context.md)"`로 system prompt 레벨 주입 가능
+
+**대안 — CLAUDE.md에서 `@import`:**
+```markdown
+# CLAUDE.md 안에서
+- @.claude/contexts/backend-context.md
+```
+이렇게 하면 system prompt 레벨로 로드되지만, **매 세션마다** 로드됩니다 (지연 로딩 아님). 항상 필요한 컨텍스트에만 사용.
+
+## 템플릿
+
+```markdown
+# [Project] Context
+
+## 기술 스택
+- Runtime: [...]
+- Framework: [...]
+- Database: [...]
+
+## API 엔드포인트
+| Method | Path | Auth | Request | Response |
+|--------|------|------|---------|----------|
+| ... | ... | ... | ... | ... |
+
+## 주요 정보
+- ...
+
+## 최근 변경 사항
+- [date]: [change]
+```
+
+## 모범 사례
+
+1. 컨텍스트 파일은 100줄 이내로 유지
+2. 주요 변경 사항 후 업데이트
+3. 비밀 정보 포함 금지 (어차피 삭제됨)
+4. Claude가 알아야 할 **참조 데이터**에 집중
+5. **행동 규칙**은 `.claude/rules/`에 넣기 — 더 높은 지시 우선순위
+6. 세션당 2개 이상의 컨텍스트 로드 지양

package/.claude/contexts/README.md

@@ -0,0 +1,94 @@
+> **[한국어 버전](README.ko.md)**
+
+# Contexts Directory
+
+## Purpose
+Contains **project reference data** for lazy loading. These are NOT loaded automatically—use `/load-context` to load them on demand.
+
+Context files are static project information (API endpoints, DB schema, env vars). They differ from sessions, which track in-progress work state.
+
+| | Contexts | Sessions |
+|--|----------|----------|
+| **Nature** | Static project reference data | In-progress work state |
+| **Changes** | Only when project structure changes | Every session |
+| **Examples** | API endpoints, DB schema, env vars | Progress, decisions, next steps |
+| **Save needed** | No (files are already persisted) | Yes (`/session-save` each time) |
+
+## Contents
+
+| File | Purpose | When to Load |
+|------|---------|--------------|
+| `backend-context.md` | API endpoints, DB schema, env vars | Backend development |
+| `frontend-context.md` | Components, state, routes | Frontend development |
+
+## Usage
+
+```bash
+# Load when starting work
+/load-context backend
+
+# Check what's available
+/load-context --list
+
+# Switch focus
+/load-context frontend
+```
+
+## Why Lazy Loading?
+
+1. **Prompt Cache Efficiency:** CLAUDE.md stays immutable, maximizing cache hits
+2. **Cost Economy:** Only load what you need (Pro Plan: every message = quota)
+3. **Focused Context:** Avoid loading frontend schema during backend work
+
+## Instruction Hierarchy
+
+> **Important:** `/load-context` uses the Read tool, so content loads at **tool output level** — the lowest priority in Claude's instruction hierarchy.
+
+```
+System prompt (highest)  ← CLAUDE.md, rules/, --system-prompt
+User messages            ← Your prompts
+Tool output (lowest)     ← /load-context, @file, Read tool
+```
+
+**What this means:**
+- Context files are **reference data** Claude uses — this is fine at tool output level
+- **Strict rules** Claude must follow → put in `.claude/rules/` instead (auto-loaded, system prompt level)
+- For advanced use: `claude --system-prompt "$(cat context.md)"` injects at system prompt level
+
+**Alternative — `@import` in CLAUDE.md:**
+```markdown
+# In CLAUDE.md
+- @.claude/contexts/backend-context.md
+```
+This loads at system prompt level, but loads **every session** (no lazy loading). Use only for always-needed context.
+
+## Template
+
+```markdown
+# [Project] Context
+
+## Tech Stack
+- Runtime: [...]
+- Framework: [...]
+- Database: [...]
+
+## API Endpoints
+| Method | Path | Auth | Request | Response |
+|--------|------|------|---------|----------|
+| ... | ... | ... | ... | ... |
+
+## Key Information
+- ...
+
+## Recent Changes
+- [date]: [change]
+```
+
+## Best Practices
+
+1. Keep context files under 100 lines
+2. Update after major changes
+3. Don't include secrets (they'll be scrubbed anyway)
+4. Focus on **reference data** Claude needs to know
+5. Put **behavioral rules** in `.claude/rules/` instead — higher instruction priority
+6. Avoid loading more than 2 contexts per session

package/.claude/contexts/backend-context.md

@@ -0,0 +1,23 @@
+# Backend Context
+
+## Tech Stack
+- Runtime: [Node.js/Bun]
+- Framework: [Express/Fastify/Hono]
+- Database: [PostgreSQL/MongoDB]
+- ORM: [Prisma/Drizzle]
+
+## API Endpoints
+| Method | Path | Auth | Request | Response |
+|--------|------|------|---------|----------|
+| POST | /auth/login | No | {email,pass} | {token} |
+| GET | /users/me | Yes | - | User |
+
+## Environment
+```
+PORT=3001
+DATABASE_URL=...
+JWT_SECRET=[REDACTED]
+```
+
+## Recent Changes
+- [date]: [change]