feat: improve Cursor Agent integration structure (apply DRY principle)

- Improve structure for automatic Agent activation in Cursor
  - Add AGENTS.md: industry standard format for cross-platform compatibility
  - Add auto-agent.mdc: glob pattern-based Agent auto-activation
  - Simplify imports.mdc: reduce from 178 lines to 35 lines, core principles only
  - Update cursor.md adapter: document new integration structure
- Apply Single Source of Truth principle:
  - All Agent definitions managed only in packages/rules/.ai-rules/
  - AGENTS.md and .mdc files act as pointers only
  - Remove duplication to improve maintainability

close #147

Author: wishket-pjw

.cursor/rules/auto-agent.mdc

@@ -0,0 +1,47 @@
+---
+description: codingbuddy Agent auto-activation based on file patterns
+globs:
+  - "**/*.tsx"
+  - "**/*.ts"
+  - "**/*.go"
+  - "**/*.py"
+  - "**/*.java"
+  - "**/*.rs"
+  - "**/Dockerfile"
+  - "**/*.yml"
+  - "**/*.yaml"
+alwaysApply: false
+---
+
+# codingbuddy Agent System
+
+## Required: Mode Keyword Detection
+
+When user message starts with `PLAN`, `ACT`, `EVAL` (or localized variants):
+
+→ **Immediately** call `parse_mode` MCP tool
+
+## File Context → Agent Mapping
+
+| File Pattern | Recommended Agent | MCP Call |
+|--------------|-------------------|----------|
+| `*.tsx`, `*.ts` | frontend-developer | `get_agent_details("frontend-developer")` |
+| `*.go`, `*.py`, `*.java`, `*.rs` | backend-developer | `get_agent_details("backend-developer")` |
+| `Dockerfile`, `*.yml` | devops-engineer | `get_agent_details("devops-engineer")` |
+| `*.json` (agents/) | agent-architect | `get_agent_details("agent-architect")` |
+
+## Specialist Auto-Recommendation
+
+| Detected Topic | Recommended Specialist |
+|----------------|------------------------|
+| Security, auth, XSS, CSRF | security-specialist |
+| Accessibility, ARIA, a11y, WCAG | accessibility-specialist |
+| Performance, bundle, optimization | performance-specialist |
+| Testing, TDD, coverage | test-strategy-specialist |
+| Architecture, layers, dependencies | architecture-specialist |
+
+→ Use `prepare_parallel_agents` for parallel reviews
+
+## Reference
+
+Agent definitions: `packages/rules/.ai-rules/agents/README.md`

.cursor/rules/imports.mdc

@@ -1,177 +1,35 @@
 ---
-description: Common AI Rules Import - References shared rules from packages/rules/.ai-rules/
+description: codingbuddy common rules - applied to all conversations
 globs:
 alwaysApply: true
 ---
 
-# Common AI Rules Reference
+# codingbuddy Rules
 
-This project uses shared AI coding rules from `packages/rules/.ai-rules/` directory for consistency across all AI assistants (Cursor, Claude Code, Antigravity, Codex, Q, Kiro).
+## Workflow
 
-## 📚 Core Workflow Rules
+- **PLAN** → **ACT** → **PLAN** (default flow)
+- **EVAL** only on explicit request
 
-**Source**: [packages/rules/.ai-rules/rules/core.md](../../packages/rules/.ai-rules/rules/core.md)
+## Required Actions
 
-### Work Modes
+When `PLAN`, `ACT`, `EVAL` keywords detected → **Immediately** call `parse_mode` MCP tool
 
-You have three modes of operation:
+## Core Principles
 
-1. **PLAN mode** - Define a plan without making changes
-2. **ACT mode** - Execute the plan and make changes
-3. **EVAL mode** - Analyze results and propose improvements
+- TDD: Red → Green → Refactor
+- Test coverage 90%+
+- TypeScript strict (no `any`)
+- Server Components first
 
-**Mode Flow**:
-- Start in PLAN mode by default
-- Move to ACT when user types `ACT`
-- Return to PLAN after ACT completes (automatic)
-- Move to EVAL only when user explicitly types `EVAL`
+## Project Config
 
-**Mode Indicators**:
-- Print `# Mode: PLAN` in plan mode
-- Print `# Mode: ACT` in act mode
-- Print `# Mode: EVAL` in eval mode
+Use `get_project_config` MCP tool to retrieve project-specific settings (language, tech stack, conventions).
 
-### Agent System
+## Detailed Rules
 
-**Auto-activated Agents**:
-- **Frontend Developer** (`packages/rules/.ai-rules/agents/frontend-developer.json`): PLAN/ACT modes
-- **Code Reviewer** (`packages/rules/.ai-rules/agents/code-reviewer.json`): EVAL mode
+All details in Single Source of Truth:
 
-**Specialist Agents** (12 available):
-- Architecture, Test Strategy, Performance, Security
-- Accessibility, SEO, Design System, Documentation
-- Code Quality, DevOps Engineer
-
-For complete workflow details, see [packages/rules/.ai-rules/rules/core.md](../../packages/rules/.ai-rules/rules/core.md)
-
----
-
-## 🏗️ Project Setup
-
-**Source**: [packages/rules/.ai-rules/rules/project.md](../../packages/rules/.ai-rules/rules/project.md)
-
-### Tech Stack
-
-프로젝트의 `package.json`을 참조하세요.
-
-### Project Structure
-```
-src/
-├── app/          # Next.js App Router
-├── entities/     # Domain entities (business logic)
-├── features/     # Feature-specific UI components
-├── widgets/      # Composite widgets
-└── shared/       # Common modules
-```
-
-### Development Rules
-- **Layer dependency**: app → widgets → features → entities → shared
-- **Pure/impure separation**: Separate files for pure and impure functions
-- **Server Components**: Default, Client Components only when necessary
-- **Test coverage**: 90%+ goal
-
-For complete project setup, see [packages/rules/.ai-rules/rules/project.md](../../packages/rules/.ai-rules/rules/project.md)
-
----
-
-## 🎯 Augmented Coding Principles
-
-**Source**: [packages/rules/.ai-rules/rules/augmented-coding.md](../../packages/rules/.ai-rules/rules/augmented-coding.md)
-
-### TDD Cycle (Strict Adherence)
-
-Follow the **Red → Green → Refactor** cycle:
-
-1. **Red**: Write a failing test that defines functionality
-2. **Green**: Implement minimum code needed to pass
-3. **Refactor**: Improve structure only after tests pass
-
-### Core Principles
-- **TDD for core logic** (entities, shared/utils, shared/hooks)
-- **Test-after for UI** (features, widgets)
-- **SOLID principles** and code quality standards
-- **90%+ test coverage** goal
-- **No mocking** - test real behavior with actual implementations
-
-### Code Quality Standards
-- TypeScript strict mode (no `any`)
-- DRY (Don't Repeat Yourself)
-- Keep methods small (10-20 lines max)
-- Minimize state, prefer pure functions
-- Tidy First: Separate structural vs behavioral changes
-
-For complete augmented coding guide, see [packages/rules/.ai-rules/rules/augmented-coding.md](../../packages/rules/.ai-rules/rules/augmented-coding.md)
-
----
-
-## 🤖 Specialist Agents
-
-**Source**: [packages/rules/.ai-rules/agents/](../../packages/rules/.ai-rules/agents/)
-
-All specialist agents are defined in `packages/rules/.ai-rules/agents/` directory:
-
-| Agent | Expertise | Use Cases |
-|-------|-----------|-----------|
-| Frontend Developer | React/Next.js, TDD, design system | Component implementation, Server Actions |
-| Code Reviewer | Quality evaluation, architecture | Code review, production readiness |
-| Security Specialist | OAuth 2.0, JWT, XSS/CSRF | Authentication, security audit |
-| Accessibility Specialist | WCAG 2.1 AA, ARIA | A11y compliance, screen readers |
-| Performance Specialist | Bundle size, Core Web Vitals | Performance tuning, optimization |
-| +7 more specialists | Various domains | See agents README |
-
-For complete agent documentation, see [packages/rules/.ai-rules/agents/README.md](../../packages/rules/.ai-rules/agents/README.md)
-
----
-
-## 🔴 MANDATORY: Keyword Mode Detection
-
-<CODINGBUDDY_CRITICAL_RULE>
-
-**When user message starts with PLAN, ACT, or EVAL keyword (or localized: Korean 계획/실행/평가, Japanese 計画/実行/評価, Chinese 计划/执行/评估, Spanish PLANIFICAR/ACTUAR/EVALUAR):**
-
-1. **IMMEDIATELY** call `parse_mode` MCP tool with the user's prompt
-2. Follow the returned `instructions` **EXACTLY**
-3. Apply the returned `rules` as context
-4. If `warnings` are present, inform the user
-
-**This is MANDATORY, not optional.**
-
-Failure to call `parse_mode` when these keywords are present will result in:
-- Missed critical checklists (Devil's Advocate Analysis, Impact Radius Analysis)
-- Incomplete evaluations
-- Quality issues not caught before deployment
-
-**Red Flags** (STOP if you think these):
-| Thought | Reality |
-|---------|---------|
-| "I can handle EVAL myself" | NO. Call parse_mode FIRST. |
-| "The rules are similar anyway" | NO. Each mode has specific checklists. |
-| "I'll save a tool call" | NO. parse_mode MUST be called FIRST. |
-| "I already know what to do" | NO. Rules may have been updated. |
-
-</CODINGBUDDY_CRITICAL_RULE>
-
-Example: `PLAN design auth feature` → **즉시** parse_mode 호출 → PLAN 모드로 작업
-
----
-
-## 💬 Communication
-
-- **Always respond in Korean (한국어)**
-- User frequently modifies code directly, so **always read code and refresh information** instead of relying on memory
-- **Start by understanding current code state** for every question
-
----
-
-## 📖 Full Documentation
-
-For comprehensive guides:
-- **Core Rules**: [packages/rules/.ai-rules/rules/core.md](../../packages/rules/.ai-rules/rules/core.md)
-- **Project Setup**: [packages/rules/.ai-rules/rules/project.md](../../packages/rules/.ai-rules/rules/project.md)
-- **Augmented Coding**: [packages/rules/.ai-rules/rules/augmented-coding.md](../../packages/rules/.ai-rules/rules/augmented-coding.md)
-- **Agents System**: [packages/rules/.ai-rules/agents/README.md](../../packages/rules/.ai-rules/agents/README.md)
-- **Integration Guide**: [packages/rules/.ai-rules/adapters/cursor.md](../../packages/rules/.ai-rules/adapters/cursor.md)
-
----
-
-**Note**: This file references common AI rules from `packages/rules/.ai-rules/` directory. All AI assistants (Cursor, Claude Code, Antigravity, etc.) share the same rules for consistency.
+- [`packages/rules/.ai-rules/rules/core.md`](../../packages/rules/.ai-rules/rules/core.md)
+- [`packages/rules/.ai-rules/rules/augmented-coding.md`](../../packages/rules/.ai-rules/rules/augmented-coding.md)
+- [`packages/rules/.ai-rules/agents/README.md`](../../packages/rules/.ai-rules/agents/README.md)

AGENTS.md

@@ -0,0 +1,80 @@
+# AGENTS.md
+
+This project uses [codingbuddy](https://www.npmjs.com/package/codingbuddy) MCP server to manage AI Agents.
+
+## Quick Start
+
+### For End Users (Your Project)
+
+1. **Install MCP Server**
+
+   ```json
+   // Claude Code: .claude/settings.json
+   {
+     "mcpServers": {
+       "codingbuddy": {
+         "command": "npx",
+         "args": ["-y", "codingbuddy"]
+       }
+     }
+   }
+   ```
+
+   ```json
+   // Cursor: .cursor/mcp.json
+   {
+     "mcpServers": {
+       "codingbuddy": {
+         "command": "npx",
+         "args": ["-y", "codingbuddy"]
+       }
+     }
+   }
+   ```
+
+2. **Use Mode Keywords**
+
+   ```
+   PLAN Design a login feature
+   ```
+
+   → `parse_mode` tool is automatically called, loading appropriate Agent and rules.
+
+### For Monorepo Contributors
+
+Direct file references work within this repository:
+
+- Workflow: `packages/rules/.ai-rules/rules/core.md`
+- Coding principles: `packages/rules/.ai-rules/rules/augmented-coding.md`
+- Agent list: `packages/rules/.ai-rules/agents/README.md`
+
+## Workflow Modes
+
+| Keyword | Mode | Description |
+|---------|------|-------------|
+| `PLAN` | Plan | Design and planning before implementation |
+| `ACT` | Act | Implement code following the plan |
+| `EVAL` | Eval | Code quality review and improvement |
+
+Multi-language keywords supported:
+- 한국어: `계획`, `실행`, `평가`
+- 日本語: `計画`, `実行`, `評価`
+- 中文: `计划`, `执行`, `评估`
+- Español: `PLANIFICAR`, `ACTUAR`, `EVALUAR`
+
+## MCP Tools
+
+| Tool | Purpose |
+|------|---------|
+| `parse_mode` | Parse mode keywords, load Agent and rules |
+| `get_agent_details` | Get specific Agent details |
+| `get_project_config` | Get project configuration |
+| `recommend_skills` | Recommend skills based on prompt |
+| `search_rules` | Search across all rules |
+
+## Reference
+
+- **Rules**: [`packages/rules/.ai-rules/rules/`](packages/rules/.ai-rules/rules/)
+- **Agents**: [`packages/rules/.ai-rules/agents/`](packages/rules/.ai-rules/agents/)
+- **Skills**: [`packages/rules/.ai-rules/skills/`](packages/rules/.ai-rules/skills/)
+- **MCP API**: [`docs/api.md`](docs/api.md)