code-reviewer

Files (1)

• SKILL.md

SKILL.md (first lines)

# Role
Review Agent

# Trigger
- User invokes REVIEW workflow.
- Router routes REVIEW intent to code-reviewer.

# Description
The Code-Reviewer skill provides systematic code review with confidence scoring. It performs two-stage reviews (spec compliance → quality), reports only issues with ≥80 confidence, and identifies CRITICAL issues that block shipping.

# Agent Capabilities
- **Two-Stage Review:** Spec compliance first, then quality analysis
- **Confidence Scoring:** Quantify issue confidence based on evidence
- **Severity Classification:** CRITICAL/IMPORTANT/MINOR categorization
- **Git Context:** Use recent changes and blame to focus review
- **Structured Reporting:** Generate comprehensive review reports
- **Blocking Decisions:** Identify issues that prevent shipping

# Inputs
- User request for review
- Code files or PR changes
- Git context (recent commits, blame)
- Requirements/specification documents
- Focus areas (security, quality, performance, all)

# Outputs
- Review report
- Issues found (with confidence scores)
- Router contract (YAML)

# Steps

## 1. Clarify Scope (if needed)

If scope is unclear, ask user for:
- Files to review
- Focus areas (security, quality, performance, all)
- Blocking threshold (all issues or only CRITICAL)

**Scope Clarification Questions:**
- "Which files should I review?"
- "What are your focus areas? (security, performance, quality, all)"
- "Do you want to see all issues or only blocking ones?"

## 2. Gather Git Context

Collect context about recent changes:
```
git_context = {
    "recent_commits": [...],
    "changed_files": [...],
    "blame_info": {...},
}
```

**Git Context Checklist:**
- [ ] Recent commits affecting files
- [ ] Changed files in PR
- [ ] Blame for changed lines
- [ ] Related issues or PRs

## 3. Stage 1: Spec Compliance Review

Verify code meets requirements:

**Checklist:**
- [ ] All required features implemented
- [ ] Behavior matches specification
- [ ] Edge cases handled per spec
- [ ] Error conditions handled per spec
- [ ] API contracts satisfied
- [ ] Data models match requirements

**Output for Stage 1:**
```yaml
spec_compliance:
  meets_requirements: true | false
  missing_features: []
  incorrect_behavior: []
  compliant_items: []
```

## 4. Stage 2: Code Quality Review

Analyze code quality across dimensions:

### Security Analysis
- [ ] Injection vulnerabilities (SQL, XSS, command)
- [ ] Authentication/authorization issues
- [ ] Sensitive data exposure
- [ ] Cryptography usage
- [ ] Input validation

### Performance Analysis
- [ ] N+1 query patterns
- [ ] Inefficient algorithms
- [ ] Memory leaks
- [ ] Unnecessary allocations
- [ ] Blocking operations

### Code Quality Analysis
- [ ] Readability and maintainability
- [ ] Naming conventions
- [ ] Function complexity
- [ ] Error handling
- [ ] Test coverage

### Best Practices
- [ ] Design patterns
- [ ] SOLID principles
- [ ] DRY principle
- [ ] Documentation

**Output for Stage 2:**
```yaml
code_quality:
  security_issues: []
  performance_concerns: []
  maintainability_issues: []

From lshtram/nso, MIT.

More from lshtram/nso