🚀 AI Code Reviewer Plus

AI-powered intelligent code reviewer — multi-dimensional semantic analysis, five-tier severity classification, actionable fix suggestions via Claude Code Skill.

中文文档 | 完整文档

🎯 The Problem It Solves

Traditional linters catch syntax errors, but miss semantic issues:

// Linter: ✅ No syntax errors
function processUser(input: any) {
  return input.name.trim()  // Runtime: 💥 if input is null
}

Questions linters can't answer:

• Is this business logic correct?
• Are there security vulnerabilities?
• Will this cause performance issues?
• Is this maintainable?

| Scenario | Traditional Linter | AI Code Reviewer | |----------|-------------------|------------------| | "Is this safe?" | Pattern matching, can't understand context | AI understands business semantics | | "How to fix?" | Reports error only | Provides executable fix snippets | | "Is it critical?" | Binary pass/fail | 5-tier severity: BLOCKER → SUGGESTION |

✨ The Solution

AI Code Reviewer uses semantic understanding across five dimensions:

| Dimension | Focus | Example Rules | |-----------|-------|---------------| | Correctness | Logic errors, null pointers | COR-001: Null check missing | | Security | XSS, SQL injection, data leakage | SEC-001: User input unsanitized | | Performance | N+1 queries, memory leaks | PER-001: Loop inside loop | | Maintainability | Duplicate code, complexity | MAIN-001: Duplicated logic | | Best Practices | Framework conventions | BP-001: Vue reactive rules |

📊 Five-Tier Severity

Not all issues are equal. AI Code Reviewer classifies by impact:

| Level | Icon | Meaning | Action | |-------|------|---------|--------| | BLOCKER | 🚫 | Must fix, blocks merge | Fix immediately | | HIGH | 🔴 | Critical, breaks functionality | Fix this iteration | | MEDIUM | 🟡 | Important, affects quality | Plan to fix | | LOW | 🟢 | Minor, polish level | Fix later | | SUGGESTION | 💡 | Optional improvement | Optional |

✨ Core Features

🔍 Multi-Dimensional Analysis

• Correctness — Logic errors, null pointers, boundary conditions
• Security — XSS, SQL injection, data leakage
• Performance — N+1 queries, memory leaks, unnecessary re-renders
• Maintainability — Duplicate code, complexity, naming
• Best Practices — Framework conventions, code style

📊 Five-Tier Severity Classification

| Level | Meaning | Action | |-------|---------|--------| | 🚫 BLOCKER | Must fix, blocks merge | Fix immediately | | 🔴 HIGH | High priority | Fix this iteration | | 🟡 MEDIUM | Medium priority | Plan to fix | | 🟢 LOW | Low priority | Fix later | | 💡 SUGGESTION | Optional improvement | Optional |

🚀 Getting Started

Option 1: Claude Code Plugin (Recommended)

# In Claude Code, run:
/plugin marketplace add saqqdy/ai-code-reviewer-plus
/plugin install ai-code-reviewer-plus

Available Commands

| Command | Description | |---------|-------------| | /review [branch] | Review diff against target branch | | /review-file <file> | Review a single file | | /review-commit <hash> | Review a specific commit |

Option 2: CLI (Zero-Install)

npx ai-code-reviewer-plus review --branch main
npx ai-code-reviewer-plus review-file src/index.ts
npx ai-code-reviewer-plus review-commit abc1234

Option 3: Programmatic Usage

pnpm add ai-code-reviewer-plus

import { collectDiff, detectProject } from 'ai-code-reviewer-plus'

const diffs = await collectDiff({
  root: process.cwd(),
  targetBranch: 'main'
})

const project = await detectProject(process.cwd())

📋 Example Output

/review main

🔍 Reviewing diff against main...

📁 Project: vue3 (typescript)
📝 Files changed: 3

📊 Review Summary:
- Total findings: 12
- 🚫 Blockers: 1
- 🔴 High: 3
- 🟡 Medium: 5
- 🟢 Low: 2
- 💡 Suggestions: 1

🚫 Blockers (Must Fix):

- 🚫 **SEC-001** [BLOCKER] `src/auth.ts:45`
  **XSS vulnerability in user input**
  
  User input is directly rendered without sanitization.
  ```typescript
  element.innerHTML = userInput  // 💥 Dangerous!

💡 Fix: Sanitize user input before rendering.

element.textContent = sanitize(userInput)

Confidence: 🟢 High

---

## 🚀 Quick Experience

Not sure where to start? Follow this path:

| Step | Method | Time | What You'll See |
|------|--------|------|-----------------|
| 1 | CLI | 1 min | Diff collection + project detection |
| 2 | API example | 3 min | Programmatic interface + formatting |
| 3 | Skill | 5 min | **Full review** (rules + AI analysis) |

**Fastest path**: Install the Skill, run `/review main` — one command, full experience.

```bash
# Step 1: Try CLI instantly (in any git repo)
npx ai-code-reviewer-plus review --branch main

# Step 2: Run the API example
git clone https://github.com/saqqdy/ai-code-reviewer-plus.git
cd ai-code-reviewer-plus && pnpm install
npx tsx examples/basic-usage.ts

# Step 3: Full AI-powered review (in Claude Code)
/plugin install ai-code-reviewer-plus
/review main

⚙️ Configuration

Create .ai-code-reviewer-plus.yml to customize:

rules:
  enabled:
    - COR-001  # Null pointer check
    - SEC-001  # XSS prevention
    - SEC-002  # SQL injection
    - PER-001  # N+1 query detection
  disabled:
    - STYLE-001  # Naming conventions
  severityOverrides:
    SEC-001: BLOCKER  # XSS must fix immediately

excludePaths:
  - node_modules/
  - dist/
  - coverage/

maxFindingsPerFile: 20
outputFormat: markdown

🗂️ Project Structure

ai-code-reviewer-plus/
├── .claude/skills/ai-code-reviewer-plus/  # Claude Code Skill
├── src/                               # TypeScript source
│   ├── collectors/                    # Git data collectors
│   ├── utils/                         # Utilities
│   └── types.ts                       # Type definitions
├── docs/                              # VitePress docs
└── examples/                          # Usage examples

🛠️ Development

pnpm install          # Install dependencies
pnpm run lint         # ESLint check
pnpm run typecheck    # TypeScript check
pnpm run test         # Run tests
pnpm run build        # Build (ESM + CJS)
pnpm run docs:dev     # Start docs server

📋 Version Roadmap

| Version | Theme | Status | |---------|-------|--------| | v0.1.0 | Core framework + CLI + Skill | ✅ Current | | v0.2.0 | Rule library + AI analysis engine | 📋 Planned | | v1.0.0 | Production-ready + Marketplace | 📋 Planned |

📄 License

MIT