multi-agent-audit

You can run a comprehensive code audit by orchestrating seven specialized agents across three parallel waves, covering code quality, security, performance, architecture, and compliance. This multi-agent audit surfaces convergent findings in a single pass and consolidates them by severity.

Automating parallel code review is achieved by dispatching multiple agents in coordinated waves to analyze diverse bug classes concurrently. The three-wave orchestration covers various vulnerability types and generates a unified findings document for traceability.

Yes, you can restrict the scope of a multi-agent security audit to specific directories. The audit supports scope and output path control, allowing you to target specific codebases while maintaining read-only analysis.

A bug reporting system is not required to use the multi-agent audit findings, but it is optionally supported. When enabled, the audit writes consolidated findings to the bug reports database for traceability; otherwise, it outputs results to the .claude/audits directory.

Multi-agent audit consolidation handles severity ranking by aggregating findings from all seven agents after the three waves complete. It produces a unified findings document that categorizes and ranks issues by severity for prioritized remediation.

A key limitation is that parallel agents for code quality and compliance checks require explicit prompts and operate in read-only analysis mode. They cannot modify code directly and depend on your prompt clarity to scope the audit boundaries effectively.