auto-review-loop-minimax

Autonomous iterative review loops are automated by triggering the MiniMax API to obtain external feedback, implement fixes, and re-review until a positive assessment or maximum rounds are reached. The process persists state to REVIEW_STATE.json and logs full responses to AUTO_REVIEW.md.

Yes, MiniMax API external review supports dual backend capability. It can use an MCP tool when available, or automatically fall back to curl-based MiniMax API requests to maintain the autonomous review loop.

Round-state persistence saves the current progress of multi-round research review cycles to a REVIEW_STATE.json file. This records the phase and iteration count, allowing the autonomous loop to resume accurately after implementation fixes.

Autonomous research review loops stop automatically when the external reviewer provides a positive assessment or when the iteration reaches the defined MAX_ROUNDS limit, preventing infinite cycles and ensuring controlled iterative feedback.

No external dependencies are required to function. The automated review loop can operate using standard curl commands as a backend to query the MiniMax API for external feedback if no MCP tool integration is available.

Full reviewer responses are saved to AUTO_REVIEW.md to provide a persistent logging mechanism. This ensures that all external feedback and iterative assessments from each round are documented for later analysis in the research workflow.