sg-code-audit

What problem does it solve?

ShipGuard's sg-code-audit automates the tedious and error-prone task of auditing a codebase for bugs, security issues, and integration mismatches by dispatching parallel agents that inspect, report, and optionally fix problems across the repository. It reduces manual review time, surfaces regressions introduced by changes, and produces structured, machine-readable findings for review and CI integration.

Core Features & Use Cases

• Parallel, zone-based auditing: split a repo into non-overlapping zones and run concurrent agents to scale reviews across large codebases.
• Modes and scopes: support quick, standard, deep, and paranoid modes plus diff-scoped or full-codebase runs for targeted or comprehensive audits.
• Structured outputs & monitoring: write per-zone JSON results consumable by a review dashboard, and optionally serve a local monitor for live progress.
• Model & verification strategies: choose model strategies, enforce severity/category schemas, and include multi-round depth passes for runtime and edge-case analysis.
• Use case: run on a feature branch to audit only changed files and their importers, then send aggregated JSON to the visual review dashboard for triage and human validation.

Quick Start

Run /sg-code-audit in the conversation to scan the repository, produce zone JSON results, and optionally apply fixes.