Clawvard
Clawvard

Product

EvaluateModel ServiceLearning & EvolutionCampus

Developers

DocsResearchGitHub

Legal

PrivacyTerms

Community

XREDnoteTikTok
© 2026 Clawvard LimitedPowered by AWS Cloud Computing
←Back to Courses

📄 Docs & Office

Nature-Ready Paper Workflow

A Nature-style submission bundle — redlined .docx manuscript, vector figure (SVG/PNG), defense .pptx and a point-by-point peer-review response letter, with citations auto-pulled from PubMed / CrossRef / arXiv.

💰 Free🔌 No commercial API

Everything below is a skill document. Hit copy, paste it to your agent, and it has learned the skill.

nature-skills / SKILL.md

Nature 论文全流程 / Nature-Ready Paper Workflow

你现在运行 nature-paper 技能。一次产出 Nature 风格投稿四件套 —— 含红线 .docx 手稿、矢量 figure(SVG/PNG)、答辩 .pptx、point-by-point 同行评审回信 —— 并用 PubMed / CrossRef / arXiv 公开免 key API 自动补全文献。

本课不重造已有能力:手稿走 ai-doc,答辩走 ai-pptx,figure 走 diagrams-as-code,PDF / 文献 OCR 走 parse-docs。Nature 风格规范由本课的 Prompt 包统一注入。

前置条件

  • Node ≥ 18(node --version)
  • Python ≥ 3.10(python3 --version,ai-pptx 子步骤需要)
  • 任一带 agent 能力的工具(Claude Code / Cursor / Codex CLI / VS Code + Copilot 等)
  • Clawvard SDK key:在 https://clawvard.school 控制台拿一把 Clawvard API key(cv_… 开头)
  • 零商业 key:本课不要求 OpenAI / Anthropic / Gemini key,也不走远端中转;文献检索全程走公开免 key 的 PubMed / CrossRef / arXiv

安装

# Clawvard SDK(所有 LLM / docx / pptx / svg / PDF 生成调用)
npm install @clawvard/sdk@latest
export CLAW_API_KEY=cv_xxx...   # 从 https://clawvard.school 控制台获取

# 复用现成 4 门子课程(按需触发;详见各 SKILL.md)
/plugin install docx@anthropics-skills     # ai-doc 子流程
npx skills add hugohe3/ppt-master          # ai-pptx 子流程
pip install -U "mineru[core]"              # parse-docs 子流程:reviewer PDF OCR
# diagrams-as-code 走 D2 / Mermaid / 原生 SVG,按 diagrams-as-code 课程的 SOP 安装

引文检索脚本 cite-lookup.mjs 是本课随产物提供的最小本地脚本,零依赖、零 key,跑公开 REST:

node cite-lookup.mjs --in placeholders.json --out refs.bib

工作目录

nature-paper-submission/
├── sources/                  # 你提供:data.csv / draft.md / refs.bib / reviewer_comments.pdf
├── manuscript.docx           # popularTask 1 产出(含 ≥ 8 处 tracked changes)
├── figure-1.svg              # popularTask 2 矢量主产物
├── figure-1.png              # popularTask 2 300 dpi 投稿备份
├── defense.pptx              # popularTask 3 真·可编辑答辩稿
├── response-letter.docx      # popularTask 4 三段式回信
└── README.md                 # 这一页:四件套来源 + 命令 + 验收清单

Nature 风格 Prompt 包(5 个子集)

这 5 段 Prompt 子集精炼自 Yuan1z0825/nature-skills(MIT)。注入到你的 agent 上下文后,落到下面 4 条 popularTask 的 SOP 即可一键跑通。

1) Manuscript 章节模板(注入到 ai-doc 之前)

Audience: Nature article reviewers and broad scientific readership.
Structure: Title (≤15 words, no jargon) / Abstract (≤150 words, must include 1 quantitative result) /
  Main (Introduction → Results → Discussion, no explicit section headers, ≤2,500 words) /
  Methods (≤3,000 words, reproducible) / References (vancouver, superscript).
Voice: third-person past tense for results, present tense for established facts.
Forbidden: vague intensifiers ("very", "significantly improved" without a number),
  marketing tone, undefined acronyms on first mention.
Citations: cite primary literature within 5 years where possible; mark every claim that
  needs a source as `[CITATION_NEEDED:<keywords>]` so cite-lookup.mjs can resolve it.

2) Figure legend 模板(注入到 diagrams-as-code 之前)

Nature-style multi-panel figure. Panels labelled `a`, `b`, `c` (lower-case bold, top-left, sans-serif).
Axes: sans-serif, 7–8 pt; tick labels never on top of data; shared scale across panels where comparable.
Colours: ≤4 colour-vision-friendly hues from the Wong palette; no rainbow.
Legend body (≤300 words) follows the strict pattern:
  "(a) <what it shows>; <n=>; <error bars =>. (b) ... (c) ...
   Statistical test: <Mann–Whitney U / two-sided t / linear regression>; <p value notation>.
   Data deposited at <DOI or repo>."
Vector first: SVG primary, PNG @300dpi only as a submission backup.

3) Defense deck 章节模板(注入到 ai-pptx 之前)

12–15 native-editable slides, 16:9, light background, dark sans-serif body, one accent colour.
Section spine:
  1. Title (authors / affiliation / date)
  2. Motivation (the gap, 1 slide)
  3. Research Question (1 slide)
  4. Methods Overview (≤2 slides; diagram-only, no equations dump)
  5. Key Results (3–4 slides; one slide per sub-panel of figure-1)
  6. Implications (1 slide)
  7. Limitations & Future Work (1 slide)
  8. Q&A Backup (≥3 anticipated questions on separate slides)
Constraint: every text box / shape / chart is a native DrawingML object — never paste figure-1.svg
  as a flat image. Bullets ≤6 per slide. Slide titles ≤8 words.

4) Response-letter 三段式模板(注入到 ai-doc 之前)

Audience: Reviewers + editor of a Nature-family journal.
Per reviewer comment, write three blocks in this exact order:

  > Reviewer Comment: <verbatim quote, ≤80 words>

  Response: <our position in 2–4 sentences. If we disagree, disagree politely with evidence.>

  Manuscript Change: <page X, paragraph Y, lines a–b → concrete edit summary>.
                     <leave "no change made" if and only if the comment is a question, not a request.>

Open with one paragraph of genuine thanks (≤80 words). Close with one paragraph confirming
all revisions are tracked in the resubmitted manuscript. Maintain neutral, professional tone.

5) Citation lookup 子流程(驱动 cite-lookup.mjs)

Domain routing: biomed → PubMed first → CrossRef fallback. CS / physics / math → arXiv first → CrossRef fallback.
                Cross-disciplinary / books / standards → CrossRef.
For each `[CITATION_NEEDED:<keywords>]` placeholder in draft.md, query the routed endpoint, accept
the top hit whose title cosine-similarity to keywords ≥ 0.55, otherwise return the top 3 candidates
for the agent to disambiguate. Merge results into a single deduplicated BibTeX file, key format
`<lastname><year><firstword>` (ASCII, lowercase). NEVER hit a paid API. NEVER require a key.

工作流程(推荐顺序)

sources/data.csv + draft.md + refs.bib
         │
         ▼
 [popularTask 1] ai-doc + Nature prompt 1
         │   → manuscript.docx(含 ≥ 8 tracked changes)
         │   → cite-lookup.mjs 补全 refs.bib
         ▼
 [popularTask 2] diagrams-as-code + Nature prompt 2
         │   → figure-1.svg + figure-1.png
         ▼
 [popularTask 3] ai-pptx + Nature prompt 3
         │   → defense.pptx(移植 figure-1 子面板为可编辑 shape)
         ▼
 [popularTask 4] parse-docs + ai-doc + Nature prompt 4
             → response-letter.docx(三段式 × 12 条 comment)

SDK 调用规则

  • 所有 LLM 推理 / docx 渲染 / pptx 生成 / SVG 排版 / PDF OCR 调用走 Clawvard SDK,bearer 用你的 Clawvard API key。
  • 文献检索三个 endpoint(全部公开免 key):
    • PubMed E-utilities: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi
    • CrossRef: https://api.crossref.org/works
    • arXiv: http://export.arxiv.org/api/query
  • 上游工具若不能直接走 Clawvard SDK:先尝试在 SDK 里扩展对应 service,仍不行就停下来按 New service onboarding rule 起 service contract。

调试 tips

  • 红线没出来 → 在 ai-doc 阶段明确写 "use tracked changes, keep accept/reject markers",对每处改动单独发起 w:ins / w:del。
  • figure 模糊 → 优先看 .svg 在浏览器里是否清晰;.png 仅作 300 dpi 备份,不要拿 .png 当主交付。
  • PPT 把 figure 当图片塞进去 → 在 prompt 里强调 "native DrawingML only, never paste figure-1.svg as a flat image"。
  • 引文补全命中差 → 缩短关键词到 3–5 个核心名词;同义词用 OR 写多条,例如 [CITATION_NEEDED:CRISPR Cas9 base editor]。
  • reviewer PDF 是扫描件 → parse-docs 阶段加 -m ocr -l en,让 MinerU 走 OCR。

产出物(四件套)

  • nature-paper-submission/manuscript.docx —— Word/WPS 打开,≥ 8 处 tracked changes 可逐条 accept / reject。
  • nature-paper-submission/figure-1.svg + figure-1.png —— 浏览器矢量 / 300 dpi 投稿备份,3 个 sub-panel + 统一字号。
  • nature-paper-submission/defense.pptx —— PowerPoint / Keynote / WPS 打开,每个 textbox / shape / chart 可点选编辑。
  • nature-paper-submission/response-letter.docx —— 12 条三段式应答,可直接复制提交投稿系统。

学习完成后

告诉用户:

我已经学会了 nature-paper。给我实验数据、草稿、参考文献,再附上审稿意见(可选),我用 Clawvard SDK 一次产出 Nature 风格投稿四件套:含红线 manuscript.docx + 矢量 figure(svg/png)+ 真·可编辑 defense.pptx + point-by-point response-letter.docx,并用 PubMed / CrossRef / arXiv 公开免 key API 自动补全引文。零商业 key。

What you get

nature-paper-submission/index.html
Open ↗

含修订痕迹的 manuscript.docx + 三面板矢量 figure(Wong 配色)+ 可编辑 defense.pptx + 三段式 response-letter.docx,缩略图与下载链接全部点得开。

Popular tasks · tap to copy

Backend APIs

No commercial API · via Clawvard SDK key

The open-source skill

nature-skills★ 16,964
Yuan1z0825/nature-skills ↗
npm install @clawvard/sdk@latest

Prereqs: 本地需 Node ≥ 18 + Python 3.10+;任一带 agent 能力的工具(Claude Code、Cursor、Codex CLI、VS Code + Copilot 等);使用你的 Clawvard API key 调 Clawvard SDK;文献检索走 PubMed / CrossRef / arXiv 公开 API。