Codex Review is a three-tier, agent-orchestrated code quality defense designed for fast scans, deep audits, and adversarial cross-validation. It maps simpl…
Codex Review is a three-tier, agent-orchestrated code quality defense designed for fast scans, deep audits, and adversarial cross-validation. It maps simple trigger phrases to L1 quick scans (external model + agent pass), L2 deep audits (integrates the standalone bug-audit skill or built-in fallback), and L3 cross-validation with adversarial testing. Key features include read-only analysis, optional OpenAI-compatible external model (configured via CODEX_REVIEW_API_BASE, CODEX_REVIEW_API_KEY, CODEX_REVIEW_MODEL), environment-only credentials, local temp hotspot artifacts with auto-cleanup, and strict no-data-exfiltration policies. Use cases: pre-deploy gating, CI-integrated quality checks, security/privacy reviews, and developer triage workflows. Advantages: flexible depth selection, safe default operations, repeatable hotspot recording for follow-up audits, and graceful fallback when external services or bug-audit are unavailable.
This page belongs to the OpenClaw Skills learning hub with install guides, category navigation, and practical links.