Auto Review Loop Minimax

wanshuiyin/auto-claude-code-research-in-sleep

Runs autonomous research paper improvement cycles by calling MiniMax's API for external reviews, then implementing the suggested fixes until you hit a 6/10 score with a "ready" or "almost" verdict, or max out at 4 rounds. Built specifically because Codex MCP doesn't support OpenAI's Responses API with third-party providers. It persists state to JSON after each round so context compaction won't kill your loop, and it's smart enough to resume from where it left off if interrupted. Falls back to curl if the MiniMax MCP tool isn't available. The skill already loops internally, so wrapping it in another loop just burns tokens on redundant reviews without new signal.

skills.sh installs: 519
GitHub stars: 18.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add wanshuiyin/auto-claude-code-research-in-sleep --skill auto-review-loop-minimax --agent claude-code

Installs into .claude/skills of the current project.

Files

Auto Review Loop (MiniMax Version): Autonomous Research Improvement

🔒 Do not wrap this skill in /loop, /schedule, or CronCreate. Like /auto-review-loop, it already loops internally (review → fix → re-review), feeding each round's prior-round summary into the next review prompt (the backend is a stateless per-round API call, not a shared thread). An external timer re-enters from the top each tick, dropping that accumulated context and firing the verdict on wall-clock time instead of on artifact change — zero new signal, full token cost. Schedule the external wait that precedes it, not the verdict. See shared-references/external-cadence.md.

Autonomously iterate: review → implement fixes → re-review, until the external reviewer gives a positive assessment or MAX_ROUNDS is reached.