Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Agent skills for drafting, revising, auditing, and resubmitting Nature-style journal manuscripts.
| Date | Stars |
|---|---|
| 2026-07-31 | 397 |
| 2026-08-03 | 407 |
| 2026-08-06 | 407 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
<div align="center">
# 🧬 Nature-Paper-Skills
**面向 `Nature` 系列期刊稿件的 agent skill 仓库**
从初稿搭建 · 结构修订 · 图文对齐 · 引用核验 · 投稿前预检 到 返修回复的全链路
`journal-first` · `claim-driven` · 证据边界优先
<br/>
[](LICENSE)
[](docs/venue-routing.md)
[](docs/workflow-map.md)
[](docs/skill-map.md)
[](docs/installation-codex.md)
[](docs/installation-claude.md)
[](CONTRIBUTING.md)
[](https://github.com/Boom5426/Nature-Paper-Skills/stargazers)
**简体中文** · [English](README.en.md) · [快速开始](#-快速开始) · [技能地图](#-仓库里有什么) · [默认工作流](#-默认工作流)
</div>
---
> [!NOTE]
> 这是一个**强约束**的仓库,不是"通用论文写作技巧集合"。默认立场是 journal-first、claim-driven、证据边界优先;未明确 venue 时按 `Nature` 系列期刊导向处理。
## ✨ 特性
- 🎯 **一图一主张**:`figure-planner` 先定每张图的论点,`nature-figure` 出图,`figure-style` 查正确性
- 🧱 **结构先于润色**:先用 reverse outline 稳住证据链,再做句子级 `scientific-prose-style`
- 🔬 **证据边界优先**:Abstract / Introduction 不允许比下游证据更强
- 📊 **统计与图注可审计**:`stats-reporting-audit` 守住独立实验单元 n、多重比较、图注统计
- 📎 **引用卫生**:`citation-verifier` 本地扫描 + 严重度分级,先查后投
- 📦 **可直接复制**:每个 skill 自包含,脚本随目录分发,Codex 与 Claude Code 可并存
## 📦 快速开始
一条命令,不用克隆。脚本会自动识别你用的是 Codex 还是 Claude Code,装好推荐的 16 个 skill,并干净地覆盖旧版本。
```bash
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash
```
装完后彻底重启 agent 让它加载新装的 skills(退出并重新打开 Claude Code 或 Codex,不是 /clear),再把这句话发过去:
```text
用 paper-workflow 帮我判断这篇稿子下一步该用哪个 skill。
```
到这里就装完了,下面全是可选项。
<details>
<summary><b>安装选项</b></summary>
<br/>
```bash
# 不用自动识别,自己指定 agent(claude | codex | both)
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash -s -- --agent codex
# 额外装上图形技能(需要绘图后端,见下方 TIP)
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash -s -- --figure
# 只对当前项目生效,不写入 home 目录(Claude Code)
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash -s -- --agent claude --local
# 装全部 25 个 skill;或先预览,不写任何文件
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash -s -- --set all
curl -fsSL https://raw.githubusercontent.com/Boom5426/Nature-Paper-Skills/main/install.sh | bash -s -- --dry-run
```
完整参数见 `--help`。重复执行即原地升级。
</details>
<details>
<summary><b>想先看脚本内容,或者习惯从克隆装</b></summary>
<br/>
把网上的脚本直接 pipe 给 `bash` 值得警惕,这很合理。可以先读 [install.sh](install.sh),或者克隆下来本地跑:
```bash
git clone https://github.com/Boom5426/Nature-Paper-Skills.git
cd Nature-Paper-Skills
./install.sh --agent claude --figure
```
这样运行时脚本直接用你本地的克隆作为来源,不会联网下载任何东西。如果想改成拉取某个已发布版本,加 `--ref <branch|tag|sha>`。
</details>
<details>
<summary><b>想完全手动装</b></summary>
<br/>
复制整个 skill 目录,不要只复制 `SKILL.md`,因为部分 skill 自带脚本。重装前先删掉旧目录,否则上游已删除的文件会残留,同一个目录里混着两个版本。
```bash
# 安装目标:Codex 用 ~/.codex/skills;Claude Code 用 ~/.claude/skills(仅当前仓库用 .claude/skills)
DEST=~/.codex/skills
mkdir -p "$DEST"
for s in skills/core/*/ skills/venue/nature-portfolio-playbook/; do
name=$(basename "$s")
rm -rf "$DEST/$name"
cp -R "$s" "$DEST/$name"
done
```
分 agent 的细节见 [docs/installation-claude.md](docs/installation-claude.md) · [docs/installation-codex.md](docs/installation-codex.md)。
</details>
> [!TIP]
> **图形技能**(`nature-figure`、`figure-style`)默认不在推荐安装集内,因为它们需要绘图后端(Python matplotlib / seaborn 或 R ggplot2)。`nature-figure` 的可选 AI 示意图路线另需 `OPENROUTER_API_KEY`,Python / R 绘图主路线不需要。加 `--figure` 即可安装。
## 🔄 默认工作流
```mermaid
flowchart TD
A[paper-bootstrap] --> B[nature-portfolio-playbook]
B --> C["scientific-writing / manuscriptExcerpt of 8,055 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:f400db7701c71fbf, llm:Repository description: 'Agent skills for drafting, revising, auditing, and resubmitting Nature-style journal manuscripts.' Language: TeX. No topics provided.
matched fp:f400db7701c71fbf, llm:Repository description: 'Agent skills for drafting, revising, auditing, and resubmitting Nature-style journal manuscripts.' Language: TeX. No topics provided.
matched fp:f400db7701c71fbf, llm:Repository description: 'Agent skills for drafting, revising, auditing, and resubmitting Nature-style journal manuscripts.' Language: TeX. No topics provided.