Processus d'examen

model: claude-sonnet-4-0

Perform comprehensive multi-agent code review with specialized reviewers:

[Extended thinking: This tool command invokes multiple review-focused agents to provide different perspectives on code quality, security, and architecture. Each agent reviews independently, then findings are consolidated.]

Review Process

1. Code Quality Review

Use Task tool with subagent_type="code-reviewer" to examine:

• Code style and readability
• Adherence to SOLID principles
• Design patterns and anti-patterns
• Code duplication and complexity
• Documentation completeness
• Test coverage and quality

Prompt: "Perform detailed code review of: $ARGUMENTS. Focus on maintainability, readability, and best practices. Provide specific line-by-line feedback where appropriate."

2. Security Review

Use Task tool with subagent_type="security-auditor" to check:

• Authentication and authorization flaws
• Input validation and sanitization
• SQL injection and XSS vulnerabilities
• Sensitive data exposure
• Security misconfigurations
• Dependency vulnerabilities

Prompt: "Conduct security review of: $ARGUMENTS. Identify vulnerabilities, security risks, and OWASP compliance issues. Provide severity ratings and remediation steps."

3. Architecture Review

Use Task tool with subagent_type="architect-reviewer" to evaluate:

• Service boundaries and coupling
• Scalability considerations
• Design pattern appropriateness
• Technology choices
• API design quality
• Data flow and dependencies

Prompt: "Review architecture and design of: $ARGUMENTS. Evaluate scalability, maintainability, and architectural patterns. Identify potential bottlenecks and design improvements."

Consolidated Review Output

After all agents complete their reviews, consolidate findings into:

1. Critical Issues - Must fix before merge

• Security vulnerabilities
• Broken functionality
• Major architectural flaws

1. Important Issues - Should fix soon

• Performance problems
• Code quality issues
• Missing tests

1. Minor Issues - Nice to fix

• Style inconsistencies
• Documentation gaps
• Refactoring opportunities

1. Positive Findings - Good practices to highlight

• Well-designed components
• Good test coverage
• Security best practices

Target for review: $ARGUMENTS