auto-review-loop

Automated iterative review uses a secondary Codex agent to critique artifacts, automatically apply fixes, and re-evaluate outputs across multiple rounds until consensus or a maximum limit is reached. This reduces manual back-and-forth for research reports and code reviews.

An autonomous review loop is a round-robin process where a secondary agent evaluates research outputs, implements suggested fixes, and re-reviews the artifacts. It operates independently until a positive assessment is achieved or configured maximum rounds are reached.

Yes, automated review loops support configurable MAX_ROUNDS and explicit pass/fail thresholds. You can define the iteration limit and assessment criteria to control the autonomous refinement process for documentation-heavy projects and code artifacts.

Yes, the iterative review process persists detailed traceability logs and state. Results are saved to AUTO_REVIEW.md and REVIEW_STATE.json, ensuring audit readiness and allowing you to resume the automated refinement workflow if interrupted.

When max rounds are reached without a positive assessment, the automated review loop triggers safe fallbacks. The process terminates gracefully while persisting the final state and logs to AUTO_REVIEW.md and REVIEW_STATE.json for manual inspection.