# code-deep-optimizer Multi-stage review-and-fix optimizer for a source file or whole repo: 18-pass audit, domain reviewer activation, Medium+ fixes applied in place, build/lint/test verification with regression back-out. A code review that ends with fixed code, not a findings list. Auto-detects languages, frameworks, and domains in the target, activates the matching reviewer skills, runs an 18-pass audit (plus an opt-in advisory track), applies every Medium+ fix in place, and verifies through the project's own build, lint, and test gates — backing out any fix that regresses them — looping until the audit comes back clean. The verification gate is the difference from a review skill: nothing ships on the model's confidence alone. A fix that breaks the build is evidence against itself and gets reverted, not rationalized. **Use it for:** "optimize this code", "run cdo", "deep code review and fix", "audit this repo and fix it until clean". **Not for:** a one-shot diff review (`/code-review`) · prose or docs (document-critique) · SQL ([deep-query-optimizer](/skills/deep-query-optimizer/)) · a prompt embedded in the code ([prompt-deep-optimizer](/skills/prompt-deep-optimizer/)).