agentsclimarketplace

Academic peer reviewer

Skill findscripter/everything-skills/09-verticals/academic-peer-reviewer

类书式 AI Agent 技能大典 · 精选/中文化/互见成网的 500+ 开源技能,可作为 Claude Code 插件市场一键安装。A curated, cross-referenced encyclopedia of 500+ open-source agent skills.

Install
npx -y skills add findscripter/everything-skills --skill academic-peer-reviewer

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

当需要审阅一篇学术论文、模拟 reviewer 反应、判断能否被接收或找出论文优缺点时使用;以批判审查→分数预测→要点精炼→正式审稿四步流程,产出含评分与建设性意见的正式 review;不适用于论文写作/润色或非学术稿件。触发词:审稿、审阅论文、review 这篇、reviewer 会怎么说、这篇能上吗、给分数、找 weakness、peer review、meta-review。

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

5.9 KB, as published. Nobody here has run it

何时使用

  • 用户要审阅一篇学术论文、模拟评审反应、判断能否被某会议/期刊接收,或系统化找出论文优缺点。
  • 需要按顶会标准(NeurIPS、ICML、ICLR、ACL、CVPR 等)给出评分预测与正式 review。
  • 需要综合多位 reviewer 意见撰写 meta-review。

不该用的边界:

  • 不用于论文写作、润色、改稿或生成(那是 paper-writing 类任务)。
  • 不用于非学术稿件(如商业方案、新闻稿)的评审。
  • 不替代真实同行评议的署名责任,仅做模拟与辅助。

步骤

四步流程:前三步用中文深度分析,第四步产出英文正式 review。可按需只执行某步或从中途开始。

  1. 批判性审查(中文):三遍读法——快速浏览(标题/摘要/引言/结论/图表,形成初印象)→ 仔细阅读(逐节读、核对数学推导、验证实验设置、检查 claim 是否被支撑)→ 批判分析(对照五维度、识别五大退稿模式、找 missing baselines/ablations/analysis)。输出基本信息、初印象、优点[S]、缺点[W]、疑问[Q]、初步判断。
  2. 分数预测(中文):对五维度各打 1-5 分,预测 Overall(1-10)与 Confidence(1-5)。整体分非简单平均——致命缺陷可直接拉低;并与目标会议接受门槛对比(clear accept / borderline / clear reject)。
  3. 要点精炼(中文):合并去重、按重要性排序、补论文引用、区分 Major / Minor、为每个缺点附改进建议。作为第四步草稿。
  4. 正式审稿(英文):把要点译写成学术英语,按 Summary / Strengths / Weaknesses / Questions / Suggestions / Minor Issues / Rating / Justification 结构输出。

指令

五大审查维度(各 1-5 分):

维度关注权重
贡献新颖性 Novelty & Contribution想法是否新?增量够大吗?
写作清晰度 Clarity & Writing非专家能否懂主贡献?符号一致吗?
实验严谨性 Experimental Rigor设计能验证 claim?有合适 baselines、误差/显著性吗?
评估完整性 Evaluation Completeness数据多样?有 ablation、error analysis、局限讨论吗?中高
方法健全性 Soundness of Method推导对吗?假设合理且明示?有无 data leakage?

五大退稿模式(识别即重点关注):

  1. 贡献不足——"This is a straightforward extension of [X]."
  2. 不清楚——"The paper is difficult to follow."
  3. 效果弱——"The improvements are marginal and may not be statistically significant."
  4. 评估不全——"Key baselines such as [X] are missing."
  5. 方法有缺——"There appears to be an error in the derivation of Eq. (X)."

Overall Score 速查:10=Strong Accept;8-9=Accept;6-7=Weak Accept/Borderline;5=Borderline;3-4=Reject;1-2=Strong Reject。 Confidence:5=本子领域专家 … 1=非我领域。

语气:评论论文不评论作者。避免"This paper is bad.";改用"The paper would benefit from…""It is unclear how…""A potential concern is…"。

示例

输入:请以 NeurIPS 2026 的标准审查这篇论文:[论文内容/PDF 路径],特别注意 reproducibility

执行:依次产出 Step 1-3 中文分析,Step 4 英文 review,例如——

## Summary
This paper proposes … (2-4 句准确复述贡献与方法)

## Strengths
1. [S1] Clear motivation and a well-designed ablation in Sec. 4.2 …

## Weaknesses
1. [W1] Key baselines such as [X, Y] are missing, making the marginal gains hard to interpret (Table 2).

## Questions for Authors
1. [Q1] How does the method behave when … ?

## Rating
- Overall: 5/10
- Confidence: 3/5

## Justification
The idea is promising but evaluation completeness is the main blocker; addressable in rebuttal.

Meta-review:粘贴三份 review,归纳共识与分歧、评估 rebuttal 是否回应主要问题、给最终建议。

注意事项

  • 优缺点必须具体到段落/公式/实验,杜绝"写得不错""实验不足"这类空泛评语。
  • 评分要与文字一致——不要给高分却写满缺点。
  • 不确定的批评写进 Questions for Authors,不作为拒稿主因。
  • Confidence 要诚实;不熟悉子领域就给低分。
  • Borderline 时:优点重大且缺点可在 camera-ready 修复 → 倾向接受;核心方法有问题 → 倾向拒稿。
  • 即使拒稿也要真诚承认优点;批评对事不对人。
  • 倫理:保密、回避利益冲突、尊重作者、建设性导向。

互见

  • fact-checking:核查论文 claim 与引用是否属实。
  • first-principles-thinking:从第一性原理判断贡献的真实新颖性与方法健全性。
  • code-reviewer:审查论文附带的开源代码与可复现性。

本条采编自 voidful/academic-skills(MIT)。

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.