audit-codebase

Audit Codebase Command

Audit the code for potential security vulnerabilities, performance issues, and adherence to best practices. Provide a detailed report with recommendations for improvements.

Process

1. Discover Codebase Structure: Use Glob and Bash to identify key directories and file types to audit.

2. Launch 3 parallel Sonnet agents to independently audit different aspects of the codebase:

   Agent #1: Security Auditor

   • Check for exposed secrets or sensitive data
   • Validate input handling to prevent injection attacks
   • Ensure proper authentication and authorization mechanisms
   • Review error handling to avoid information leakage
   • Verify use of secure libraries and dependencies
   • Return findings in format: {severity: "CRITICAL"|"HIGH"|"MEDIUM"|"LOW", file: "path/to/file.ext", line: N, finding: "...", recommendation: "..."}

   Agent #2: Performance Analyzer

   • Identify and optimize slow or inefficient code paths
   • Analyze memory usage and detect potential leaks
   • Review database queries for performance issues
   • Evaluate caching strategies and their effectiveness
   • Return findings in format: {severity: "CRITICAL"|"HIGH"|"MEDIUM"|"LOW", file: "path/to/file.ext", line: N, finding: "...", recommendation: "..."}

   Agent #3: Best Practices Reviewer

   • Ensure adherence to coding standards and style guides
   • Review documentation for completeness and clarity
   • Evaluate test coverage and effectiveness
   • Identify opportunities for code simplification and refactoring
   • Return findings in format: {severity: "CRITICAL"|"HIGH"|"MEDIUM"|"LOW", file: "path/to/file.ext", line: N, finding: "...", recommendation: "..."}

3. Aggregate Results: Collect findings from all 3 agents and merge into unified report.

4. Prioritize Findings: Sort by severity (Critical → High → Medium → Low) within each category.

Agent Invocation Examples

Use the Task tool to launch agents in parallel:

Task(subagent_type: "security-auditor", model: "sonnet", prompt: "Audit codebase for security vulnerabilities. Focus on: exposed secrets, injection attacks, auth mechanisms, error handling, dependency security. Return findings with severity levels.")

Task(subagent_type: "general-purpose", model: "sonnet", prompt: "Analyze codebase for performance issues. Focus on: inefficient code paths, memory leaks, database queries, caching strategies. Return findings with severity levels.")

Task(subagent_type: "general-purpose", model: "sonnet", prompt: "Review codebase for best practices adherence. Focus on: coding standards, documentation, test coverage, refactoring opportunities. Return findings with severity levels.")