auto-review-loop

An automated research review loop repeatedly evaluates a research artifact via Codex MCP, applies fixes, and re-reviews until a positive assessment or maximum rounds are reached, reducing manual back-and-forth.

An autonomous multi-round research review cycle performs sequential rounds of evaluation, generates improvement actions, and triggers re-review automatically until quality thresholds are met or maximum iterations are exhausted.

Research review processes can be resumed using compact recovery via REVIEW_STATE.json, which stores round history, reviewer notes, and state to enable recovery and resumption after interruption.

The automated review loop supports optional human checkpoints during evaluation rounds, allowing manual review and intervention between automated review, fix, and re-review stages.

Complex research outputs such as manuscripts, proposals, or research plans work with an automated review and improvement loop, requiring rigorous iterative quality assurance from initial review to submission readiness.

The autonomous research review loop satisfies requirements for configurable rounds, allowing users to set a maximum number of iterations before the cycle stops if a positive assessment is not reached.