Guide · Agent skill
How to use the Nature Reviewer Mock Peer Review agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
nature-reviewer is an agent skill that makes an AI agent act as a simulated journal referee. You give it a manuscript or an excerpt, and it writes reviewer-style reports on significance, validity, and major and minor concerns so you can find weak points before you submit. This guide explains what an agent skill is, how to install this one, how to ask for a review, and where its limits are.
Nature Reviewer Mock Peer ReviewYuan1z0825/nature-skills · Apache-2.0 Download the skill Quick start
- 1Download the skill ZIP. Get the nature-reviewer ZIP from this page.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for all projects, or into the project's .claude/skills/ for one project.
- 3Or upload it in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the skill.
- 4Prepare your text. Gather the full draft or the sections you want reviewed, plus a short note on the claimed contribution.
- 5Ask for a mock review. Describe the task in plain language, for example asking for a referee's view of the manuscript before submission.
- 6Read the reports and synthesis. Check each Major Concern, its evidence pointer, and whether it is marked Blocking Yes, then fix the text and rerun if needed.
What is an agent skill, and what does this one do?
An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches the skill's description. The agent does not need extra setup from you beyond installing the folder and asking for the right kind of work.
The nature-reviewer skill covers reviewer-side evaluation of scientific manuscripts or excerpts. Its description says it is for mock peer review, self-review before submission, and assessing a manuscript from a referee's point of view. It is not for drafting author rebuttals.
- Reads the manuscript or excerpt you provide, including Chinese or English text.
- Writes several independent-style reviewer reports covering significance, validity, and major and minor concerns.
- Adds a synthesis after the reports, covering consensus blocking concerns and other major concerns.
- Separates supported claims from weak or unassessable ones, and bases the review only on your text and the skill's bundled reference files.
Install the nature-reviewer skill
Download the ZIP of the skill from this page. Unzip it into one of the following locations, depending on whether you want it for all your projects or just one.
You can also install it in the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the skill there.
- Personal use across projects: unzip into ~/.claude/skills/.
- Single project: unzip into the project's .claude/skills/ folder.
- Claude apps: Settings → Capabilities → Skills, then upload the skill.
- Check that the unzipped folder contains SKILL.md at its top level, since that file defines the skill.
What does the skill need before it runs?
The skill's metadata lists no required dependencies. It does not run scripts and does not use the network, according to this page's assessment.
The only input it needs is the text you want reviewed. The more complete the material, the more confident the review. If you give only an abstract or a few sections, the skill marks the assessment boundary and says what it could not evaluate.
- Requires: nothing listed.
- Runs scripts: no.
- Uses network: no.
- Useful inputs: full draft, abstract, introduction, results, discussion, methods, figure legends, or author notes on the claimed contribution.
How do you trigger the skill with a natural request?
You do not need a special command. Describe the job in plain language and say that you want a reviewer's view of the manuscript. The skill's description lists Chinese phrases such as 模拟审稿, 投稿前自审, and 审稿人视角评估, so those work as triggers.
State the paper's claimed contribution if you have one. The skill uses that to judge significance and whether the evidence supports the claim. If you need help answering real reviewer comments, ask for that separately, because this skill does not write rebuttals.
- Example: Act as a referee for this manuscript and list the major concerns before I submit to a high-impact journal. [paste text]
- Example: 模拟审稿：请从审稿人视角评估下面这段摘要和结果，指出主要问题。 [paste text]
- Example: Assess whether the central claim in this results section is supported by the data shown. [paste text]
- Example: Give me a self-review of this draft's methods and figure legends for technical soundness.
What does the output look like?
The skill returns three mutually blind reviewer reports and one post-review synthesis, unless you ask for a different structure. Each report has an overall assessment, major strengths, Major Concerns, and Minor Comments. Major Concerns are marked Blocking Yes only when the central case cannot be established until that concern is resolved.
Each substantive concern has an ID, a claim pointer, and an evidence pointer. When the location in your text is not clear, the skill marks it as missing instead of guessing. A separate forensic consistency audit checks arithmetic, number ranges, and table-to-text agreement after the reports are complete.
- Concern IDs such as R1-M1 for major concerns and R1-m1 for minor comments.
- Blocking Yes or No for each major concern.
- A synthesis that counts a point as consensus only when at least two reviewers raised it independently.
- Concerns are not forced to a set number; if nothing is grounded at a level, the report says so.
Tips for getting useful reviews
Give the skill the actual text, not a summary of it, when you can. Claims with no matching figure, table, or method in the text will be flagged as weak or not assessable, so a complete draft produces a sharper review.
Treat the output as a self-check. It is an AI simulation grounded only in what you supplied, and it does not predict an editor's decision or whether a journal will accept the paper.
- Include figure legends and methods so the evidence pointers can be checked.
- Say which journal or audience you are targeting, since the skill evaluates against Nature-style criteria.
- Ask for a bounded review if you only have part of the paper, and read the assessment boundary first.
- Keep rebuttal drafting for a separate tool; the skill routes that work to nature-response.
What people use it for
- Stress-testing a draft's central claim before submitting to a high-impact journal.
- Checking whether figure legends and methods support the claims made in the results.
- Getting a Chinese-language self-review of an abstract before sending it to co-authors.
- Identifying barriers for non-specialist readers in a manuscript's introduction and summary.
Frequently asked questions
How do I install the nature-reviewer skill?
Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings, then Capabilities, then Skills.
Is nature-reviewer free to use?
The skill is licensed under Apache-2.0, as stated on this page. This page does not list a price, so check the licence and any terms on the source repository before relying on it.
What does nature-reviewer require to run?
The skill's metadata lists no requirements. According to this page, it does not run scripts and does not use the network. You only need to provide the manuscript text.
Can nature-reviewer predict whether Nature will accept my paper?
No. The skill says it does not claim final decisions or certainty about fit to Nature. It is a simulated referee review meant to surface weaknesses before submission.
Does nature-reviewer write rebuttals to reviewer comments?
No. The skill is for reviewer-style evaluation, not for drafting author responses. Its instructions route rebuttal work to a different skill, nature-response.
Can I use nature-reviewer in Chinese?
Yes. The skill accepts manuscript material in Chinese or English, and its description lists Chinese trigger phrases such as 模拟审稿 and 投稿前自审.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use nature-reviewer if you want a structured, evidence-pointed self-check of a manuscript from a referee's point of view before you submit, and you can supply the full text. Pick something else if you need a prediction of an editor's decision, help writing rebuttals to real reviews, or expert domain review for high-stakes work, since the output is an AI simulation based only on the material you provide.
Download the skill