what is nature-review-studio fr?

This repository packages a specialized skill for AI coding assistants like Codex, designed to simulate how Nature journal reviewers evaluate a scientific manuscript and how authors typically respond to that feedback. The skill was built by studying 1,287 real Nature peer review files from 2025 and 2026 and distilling patterns out of them, rather than being written from general assumptions about how peer review works. Installing it is a single spoken instruction to Codex asking it to install the skill from this repository's URL. The installer copies the skill files into a local folder, records the project directory it was installed for, and checks that all of its reference files are present. The whole package is small, about 11 megabytes. Once installed, it offers three functions. A review function produces a matching Word document and plain text file simulating how a set of Nature reviewers might critique a manuscript. A respond function produces the same two file formats but for a draft response to reviewers. An update function does not produce any documents, instead running an offline pipeline to refresh the underlying knowledge base. The knowledge base itself organizes reviewer concerns into twelve recurring categories such as experimental design, figures and tables, and reproducibility, each with a typical severity level drawn from how often real reviewers raised that kind of issue. It also maps six common types of manuscripts, such as pure wet lab studies or clinical cohort studies combined with machine learning, to how many reviewers Nature typically assigns and what each one tends to focus on. On the response side, it catalogs twenty-one distinct ways authors have replied to criticism in the real dataset, from acknowledging and fixing a real error to respectfully disagreeing, plus eight statuses for tracking whether each requested change has actually been completed. The project is written primarily in Python and is aimed at researchers preparing manuscripts or responses for submission to a high-profile journal, used through an AI coding assistant rather than run as a standalone application.