code-auditor-agent

audits code across multi-agent Claude workflows, flags issues before release

stdio· No auth· source

Code-auditor-agent runs multi-phase review pipelines inside Claude Code, handling both pull request audits and full codebase sweeps. PR mode chains a correctness swarm, claim verification, skeptical external review, and deduplicated security analysis. Audit mode inventories files, triages with grep, dispatches parallel discovery agents, verifies findings, fills gaps, consolidates results, and emits TODOs with an optional fix loop. Built for engineering leads and CTOs formalizing AI-assisted review on Claude Code workflows.

> what it does

Runs four-phase PR review with correctness, verification, and security passes
Executes full-codebase audits via grep triage and parallel discovery swarms
Verifies agent claims before surfacing findings to reduce hallucinated issues
Deduplicates security flags across swarm outputs to cut reviewer noise
Generates consolidated TODO lists with an optional automated fix loop