cli agents like claude code and codex can ship real apps fast — sometimes across hundreds of files you didn't read. that's the tradeoff. you shipped quickly, and now the codebase is bigger than you can hold in your head.
common claude code / codex issues we rescue: the agent made architectural decisions you didn't notice until later (and now need to unwind), features that were 90% done but the agent got confused on edge cases, tests that pass but don't actually test the thing they claim to, and codebases where the agent created three slightly different ways to do the same thing.
audit your claude code app before you launch, fix what's broken after, or get monthly maintenance — it's all handled by senior engineers using AI tools and human judgment, not scans alone.