meta_review — Claude Code prompt: research review
The meta_review prompt for Claude Code and ChatGPT assesses whether a proposed research project can be carried out. It compares related papers, ranks failure risks, and plans a low-cost test of the idea.
no license ★ 159 updated 2026-11-02
.claude/commands/meta_review.mdWhat meta_review does
The prompt writes a meta-review of a research proposal produced by another AI model. It identifies the closest papers, explains how their actual content matches or differs from the proposal, and separates claims in abstracts from those supported by the paper body.
It examines risks involving data, compute, time, and methods, then outlines a three-day experiment designed to disprove the idea. The review also includes three questions for the author. It needs the proposal, notes on related papers, and full-text excerpts from the closest papers.
What it helps with
- compare a proposal with related research
- spot gaps between an abstract and a paper's findings
- rank likely reasons a project could fail
- test a research idea with a three-day experiment
Example prompts for the agent
- Write a meta-review of this proposal with
/meta_review. - Assess whether this research project is feasible and design a three-day test that could disprove its premise.
How to install meta_review
Manually: put the files in place
.claude/commands/meta_review.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about meta_review
How do I install meta_review in Claude Code?
Add the prompt file at .claude/commands/meta_review.md.
What does meta_review do?
It assesses a research proposal's feasibility, compares it with closely related papers, and proposes an experiment to test the idea.
What does meta_review need to work?
It needs the proposal, notes on related papers, and full-text excerpts from the closest papers.