Auto Review Loop
Repeatedly gets Gemini to review research work, applies fixes, and re-reviews until it passes or hits a round limit.
- Stars
- 18,482
- Licence
- MIT
- Files
- 1 file
What it does
This skill makes an AI coding agent run an autonomous review cycle on a research project. It sends the project context to Gemini through a local MCP bridge, reads the critique and score, implements the suggested fixes, and asks again. It repeats for up to 4 rounds or until the reviewer's verdict is 'ready' or 'almost' with a score of at least 6/10. It keeps a running log file and a state file so the loop can resume after the agent's context is compacted.
Great for
- Researchers preparing a paper who want an external, cross-model critique before submission
- Codex users who want Gemini rather than a second Codex agent acting as reviewer
- Developers who want an iterative review-and-fix cycle without manual prompting each round
Skip it if
- Projects where you need a human reviewer's judgement; the loop can accept fixes the reviewer suggests without a person checking them (unless human checkpoint is enabled)
- Setups without the gemini-review MCP bridge installed and configured, since the loop cannot run its review step
- Sensitive repositories where sending full project context to an external Gemini service is not acceptable
Try asking
- “Run an auto review loop on my paper draft until the reviewer says it is ready”
- “Review until it passes, using Gemini as the reviewer, and log each round”
- “Start an autonomous iterative improvement loop on the experiments in this repo with a human checkpoint after each review”
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.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip auto-review-loop.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip auto-review-loop.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
- Enable the human checkpoint (human checkpoint: true) for the first run so you can see each round's critique before fixes are applied
- Check review-stage/AUTO_REVIEW.md after each run; the agent's fixes and the reviewer's critiques are logged there
- Make sure your project's narrative and experiment results are up to date before starting, since the reviewer only sees the context the agent provides
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.