What problem does it solve? Getting rigorous, iterative external feedback on machine learning research is slow and manual. This Skill runs an autonomous review loop that sends your research context to the MiniMax API, parses the reviewer's score and verdict, implements fixes, and re-reviews until the work is assessed as ready or a round limit is reached. ## Core Features & Use Cases - Autonomous Review Loop: Iterates review → implement fixes → re-review for up to 4 rounds, stopping early on a positive assessment (score >= 6/10 with a ready verdict). - Dual API Access: Uses the MiniMax MCP tool when available, with a curl fallback against the MiniMax chat completions endpoint. - State Persistence: Writes REVIEW_STATE.json after each round so the loop can resume after context compaction, and logs every round with verbatim reviewer responses in AUTO_REVIEW.md. - Use Case: A researcher preparing a NeurIPS submission triggers the loop on their paper claims and experiment results; the Skill collects brutal reviewer feedback, runs cheap high-impact fixes like metric additions, and produces a cumulative review log. ## Quick Start Ask the AI to run an auto review loop with MiniMax on the current research project, reviewing the claims and experiment results until the work is ready for submission.
