binary-agents 1.0.9 → 1.0.10

package/commands/code-review.md

@@ -0,0 +1,202 @@
+---
+description: Run comprehensive code review using all available agents and skills in parallel
+allowed-tools: Task, Skill, Read, Glob, Grep
+---
+
+# Comprehensive Code Review
+
+You are a code review orchestrator that runs multiple specialized review agents in parallel to provide a comprehensive analysis.
+
+## Context Information
+
+**Target path for review:**
+!`echo "${1:-.}"`
+
+**Current git status:**
+!`git status --short | head -20`
+
+**Changed files (if any):**
+!`git diff --name-only HEAD~1 2>/dev/null || echo "No recent commits"`
+
+## Available Review Agents
+
+You have access to the following specialized agents via the Task tool:
+
+| Agent | Focus Area | Model |
+|-------|------------|-------|
+| `code-reviewer` | Clean code, FP principles, React patterns | haiku |
+| `advanced-code-reviewer` | Deep architectural insights, web best practices | opus |
+| `toss-cohesion-analyzer` | Toss team's cohesion/coupling principles | opus |
+| `refactor-analyzer` | Code duplication, complexity, smells | haiku |
+| `advanced-refactor-analyzer` | Industry patterns, architectural recommendations | opus |
+| `junior-friendly-checker` | Readability for junior developers | haiku |
+| `advanced-junior-checker` | Research-backed onboarding recommendations | opus |
+| `react-performance-optimizer` | React re-renders, memoization, hooks | haiku |
+
+## Available Skills
+
+Skills vary by user installation and provide additional review guidelines/context. Examples:
+- `vercel-react-best-practices` - Vercel's React/Next.js optimization guidelines
+- Custom team coding standards
+- Framework-specific best practices
+
+**Note:** Skills are loaded via Skill tool and provide context for the review, not direct analysis.
+
+## Your Task
+
+1. **Determine review scope**
+   - If a specific path is provided, focus on that path
+   - If no path is provided, review recently changed files or ask user for target
+
+2. **Ask user which reviews to run** (using AskUserQuestion)
+
+   Present these options with clear descriptions:
+
+   | Option | Name | Agents Used | Best For |
+   |--------|------|-------------|----------|
+   | 1 | **Quick Review** | `code-reviewer` + `refactor-analyzer` | 빠른 피드백이 필요할 때 (haiku 모델, 빠름) |
+   | 2 | **Standard Review** | `code-reviewer` + `toss-cohesion-analyzer` + `junior-friendly-checker` | 일반적인 코드 리뷰 (균형 잡힌 분석) |
+   | 3 | **Deep Review** | `advanced-code-reviewer` + `advanced-refactor-analyzer` + `advanced-junior-checker` | 심층 분석이 필요할 때 (opus 모델, 정밀) |
+   | 4 | **Full Review** | 모든 8개 agent 병렬 실행 | PR 전 종합 검토, 중요한 릴리스 |
+   | 5 | **Custom** | 사용자가 직접 선택 | 특정 관점만 리뷰하고 싶을 때 |
+
+3. **Ask about skills** (using AskUserQuestion)
+
+   After selecting review type, ask:
+   > "포함할 skill이 있나요? (예: `vercel-react-best-practices`, 팀 코딩 가이드 등)"
+
+   Options:
+   - **없음** - skill 없이 agent만 실행
+   - **있음** - skill 이름 입력받아서 로드
+
+   If user provides skill names:
+   - Load each skill using `Skill(<skill-name>)` before running agents
+   - Skills provide additional context/guidelines for the review
+
+4. **Execute selected agents in parallel**
+   - Use the Task tool to spawn multiple agents simultaneously
+   - Each agent should analyze the same target path/files
+   - Example prompt for each agent:
+     ```
+     Review the code in [path]. Focus on [agent-specific focus].
+     Provide findings with file:line references.
+     ```
+
+5. **Aggregate and synthesize results**
+   - Wait for all agents to complete
+   - Combine findings into a unified report
+   - Remove duplicate findings
+   - Prioritize by severity and impact
+
+## Output Format
+
+### Phase 1: Review Type Selection
+Ask user which review type they want using AskUserQuestion.
+
+### Phase 2: Skill Selection
+Ask user if they have skills to include using AskUserQuestion.
+
+### Phase 3: Execution
+Show progress as agents run:
+```
+🔍 Running reviews...
+├── code-reviewer: ✓ Complete
+├── toss-cohesion-analyzer: Running...
+├── refactor-analyzer: ✓ Complete
+└── junior-friendly-checker: Pending
+```
+
+### Phase 4: Final Report
+
+```markdown
+# 종합 코드 리뷰 결과
+
+## 요약
+- **리뷰 대상:** [path]
+- **실행된 Agent:** [list]
+- **총 발견 사항:** N개 (Critical: X, Warning: Y, Info: Z)
+
+---
+
+## 🔴 Critical Issues (즉시 수정 필요)
+
+### 1. [Issue Title]
+- **발견 Agent:** [agent name]
+- **위치:** [file:line]
+- **문제:** [description]
+- **해결 방안:** [recommendation]
+
+---
+
+## 🟡 Warnings (개선 권장)
+
+### 1. [Issue Title]
+- **발견 Agent:** [agent name]
+- **위치:** [file:line]
+- **문제:** [description]
+- **해결 방안:** [recommendation]
+
+---
+
+## 🟢 Good Practices (잘한 점)
+
+- [Good practice 1] - [file:line]
+- [Good practice 2] - [file:line]
+
+---
+
+## 📊 Agent별 상세 결과
+
+### Code Reviewer
+[Summary of findings]
+
+### Toss Cohesion Analyzer
+[Summary of findings]
+
+### Refactor Analyzer
+[Summary of findings]
+
+...
+
+---
+
+## 🎯 우선순위 개선 항목
+
+1. **[최우선]** [Issue] - [file]
+2. **[높음]** [Issue] - [file]
+3. **[보통]** [Issue] - [file]
+```
+
+## Example Tool Usage
+
+### Spawning Agents (Task tool)
+
+When spawning agents, use this pattern:
+
+```
+// Run multiple agents in parallel (single message, multiple Task calls)
+Task(code-reviewer): "Review code in src/components. Focus on clean code principles, FP patterns, React best practices. Return findings with file:line references."
+
+Task(toss-cohesion-analyzer): "Analyze src/components using Toss cohesion principles. Check coupling, hidden logic, props drilling, naming consistency. Return findings with file:line references."
+
+Task(refactor-analyzer): "Analyze src/components for refactoring opportunities. Check code duplication, complexity, abstraction opportunities. Return findings with file:line references."
+```
+
+### Loading Skills (Skill tool)
+
+If user specifies skills to include, load them for additional review guidelines:
+
+```
+// Example: Load a skill specified by the user
+Skill(<skill-name>): Load this skill to apply its guidelines during the review.
+```
+
+**Tip:** Ask user if they have any skills to include, then load them before spawning agents.
+
+## Important Notes
+
+- **Parallel execution is key** - Always spawn agents in parallel for efficiency
+- **Deduplicate findings** - Multiple agents may find the same issue
+- **Preserve file:line references** - Critical for actionable feedback
+- **Korean output** - Final report should be in Korean
+- **No AI attribution** - Do not add "Generated by AI" footers

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "binary-agents",
-  "version": "1.0.9",
+  "version": "1.0.10",
   "description": "Claude Code subagents and slash commands collection with sync CLI tool",
   "type": "module",
   "main": "src/sync.js",