Auto Review Loop (MiniMax)
Runs up to 4 rounds of automated research review and fixes using the MiniMax API.
- Stars
- 18,482
- Licence
- MIT
- Files
- 1 file
What it does
Gives the agent a repeatable routine for improving a research project: it sends the work to an external AI reviewer (MiniMax-M3), reads the score and critique, makes fixes or runs experiments, and asks for another review. It stops when the reviewer rates the work as ready or almost ready, or after 4 rounds. It saves progress to a state file so an interrupted run can resume.
Great for
- Researchers preparing a paper who want a second opinion on draft claims and experiments before submission
- Teams that cannot use the Codex MCP reviewer and prefer MiniMax as the external reviewer
Skip it if
- Projects with confidential or unpublished data you do not want sent to a third-party API, because review prompts include research context and the reviewer is an external service
- Anyone expecting a human-quality peer review; the reviewer is a language model and its scores are advisory
Try asking
- “Run an auto review loop minimax on my paper draft in this repo”
- “minimax review of the current experiment results and the claims in the introduction”
- “Use MiniMax to review my submission and fix the top weaknesses for up to four rounds”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
auto-review-loop-minimax.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip auto-review-loop-minimax.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip auto-review-loop-minimax.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Give the skill a clear scope in the argument, such as a specific section or claim set, so the reviewer gets focused context
- Do not wrap it in /loop, /schedule or cron, as the SKILL.md warns; it already loops internally and an external timer discards accumulated context
- Check review-stage/AUTO_REVIEW.md after each run, since the full raw reviewer responses are logged there
What's inside
- SKILL.md
Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.