Your AI-generated code works. It's probably not production ready.

Short summary

AI code generation (Claude Code, Cursor) ships fast but typically fails in production on auth, secrets, caching, and testing. The bottleneck has shifted from writing code to reviewing and hardening it. A six-part audit—auth flows, secrets, injection risks, code quality, performance, and compliance—identifies gaps before they become incidents.

• •AI-generated code excels at implementation but misses production edge cases: misconfigured auth, exposed secrets, missing caching, weak testing, slow queries, and unhandled failures
• •Code review and hardening now require different skills than code generation; most teams lack formal processes for vetting AI output before shipping
• •Six-part audit checklist (auth, secrets/injection, code quality, performance, compliance, observability) separates prototype code from production-ready systems

Generated with AI, which can make mistakes.