📖 《把 AI 当助手——经济实证研究的诚信、方法与工具》双语网页版:每章中文部在前,English Edition 在后,内容与最新合订 PDF 一致。

附录 E · MCP / Skill / CLI 工具速查矩阵

---

E.1 MCP 服务器（连接层）

E.1.1 文献检索 / 获取（第 3 章核心）★★★

表 E.1 文献检索 / 获取（第 3 章核心）速查表。

L4|L4|L4|L4

MCP 名称 & 工具数 & 核心工具 & 对应章节
nber-cli-mcp & 3 & search_papers / get_paper_info / download_paper（支持日期过滤与分页） & 第 3 章
scansci-pdf & 23 & / / （PDF 直推 Zotero）/ / / / tor 支持 & 第 3 章
zenodo-mcp & 5 & search_records / get_metadata / get_citation（直接给引用格式）/ detect_data_type / list_files & 第 3 / 6 章\

E.1.2 国际组织数据（第 6 章数据来源全景）★★

表 E.2 国际组织数据（第 6 章数据来源全景）。

L4|L4|L4|L4

MCP 名称 & 工具数 & 核心工具 & 适用场景
imf-mcp-server & 5 & imf_list_databases / imf_get_database / imf_query_dataset / imf_dataframe_query & 国际宏观金融（IFS / WEO）
oecd-mcp & 9 & list_dataflows / search_dataflows / query_data / search_indicators / get_dataflow_url & 发达经济体 + 农业政策库
faostat-mcp-server & 6 & / / / & FAO 69 域农业数据（农经核心）\

[提示] 选型口诀：农业问题找 FAO、国际比较找 IMF、发达经济体找 OECD。

E.1.3 计量与写作（第 1 / 8 / 9 / 11 章）★★

表 E.3 计量与写作（第 1 / 8 / 9 / 11 章）速查表。

L4|L4|L4|L4

MCP 名称 & 工具数 & 核心工具 & 对应章节
stata-mcp & 3 & stata_session / stata_run_file / stata_run_selection（官方端口已配置） & 第 1 / 8 / 9 章
officecli-mcp & 1 & officecli（Word / PPT / Excel 全能编辑；写前 load_skill word academic-paper） & 第 1 / 8 / 11 章
mcp-server-chart & 27 & 27 种图表：散点 / 桑基 / 思维导图 / 甘特 / 雷达 / 韦恩 / 词云 / 漏斗等 & 第 8 章
sequential-thinking & 1 & sequentialthinking（结构化思维链，论证分步推理） & 第 4 章\

E.1.4 金融数据（选用，一带而过）

表 E.4 金融数据（选用，一带而过）。

L3|L3|L3

MCP 名称 & 工具数 & 覆盖范围
tushareMcp & 254 & A 股 / 基金 / 期货 / 期权 / 债券 / 宏观
china-stock-mcp & 30 & A 股 + 板块 + 财务 + 研报
hexin-ifind-ds-* 系列 & 7 个子服务，35+ 工具 & stock / bond / fund / edb / global-stock / news / index\

E.1.5 网页交互（第 3 章拓展）

表 E.5 网页交互（第 3 章拓展）。

L3|L3|L3

MCP 名称 & 工具数 & 用途
browser-use & 16 & 出版社站点、数据库平台交互
chrome-devtools & 29 & Chrome 调试、take_screenshot、evaluate_script DOM 审计\

△  CNKI 边界：本手册不展开 cnki-mcp（手册 v1.4 锁定）；中文文献请用知网网页自行检索，配合 Zotero jasminum 插件抓取元数据（第 3 章 §3.5）。

---

E.2 Skill（知识层）

E.2.1 文献检索 / 综述组（10 个）★★★

表 E.6 文献检索 / 综述组（10 个）速查表。

L3|L3|L3

Skill & 描述要点 & 对应章节
arxiv & arXiv 搜索 / 下载 / 摘要，保存到本地论文库 & 第 3 章
litreview & Consensus 检索 + PICO/SPIDER 框架 + Word 研究指南（“launching pad”） & 第 3 章
find-literature & “find papers / what does the literature say / cite papers” & 第 3 章
paper-finder-research & Allen AI PaperFinder 智能文献综述 + 自动下载保存 & 第 3 章
research-lit & 论文搜索 / 分析 / 相关工作 / key ideas & 第 3 章
literature-toolkit & Academic literature CLI: search, parse, and analyze & 第 3 章
lit-review-assistant & 经济学文献综述助手（Search, summarize, synthesize） & 第 3 章
systematic-literature-review & 多源检索 → 去重 → AI 评分 → 选文 → 专家写作，PDF + Word，多语言 & 第 3 章
academic-literature-guide & 4 层递进导读：直觉 → 概念 → 技术 → 批判 & 第 3 章
academic-paper-analysis & PDF 深度分析提取结构化 Markdown，含写作与论证解析 & 第 3 / 10 章\

E.2.2 研究全流程组（7 个）★★

表 E.7 研究全流程组（7 个）。

L3|L3|L3

Skill & 描述 & 对应章节
academic-researcher & Systematic academic research from topic to literature synthesis & 第 2 / 3 章
research-ideation & Generate research questions from economic phenomena & 第 2 章
research-refine & 方向细化（GPT 迭代评审，problem-anchored） & 第 4 章
research-review & GPT 深度批判性评审（via Codex MCP） & 第 4 / 11 章
research-pipeline & Workflow 1 + 2 全流程编排 & 课外
Academic-Research-Agent-Skill & 18 命令框架：scope → reality-gate → novelty-review → claim verification & 课外（github 高星）
Awesome-Vibe-Research & 科研全流程方法论文档（8 阶段） & 课外（github 高星）\

E.2.3 论文写作组（12 个）★★

表 E.8 论文写作组（12 个）。

L3|L3|L3

Skill & 描述 & 对应章节
academic-paper-writer & Draft economics papers with proper structure and academic style & 第 11 章
paper-plan & 论文大纲（review conclusions → outline） & 第 11 章
paper-write & LaTeX 分章起草 & 第 11 章
paper-figure & 论文图表生成 & 第 8 / 11 章
paper-compile & LaTeX 编译 + 修复 + 验证 & 第 11 章
paper-writing & Workflow 3 全流程编排（plan → figure → write → compile → 润色循环） & 第 11 章
auto-paper-improvement-loop & GPT 评审 → 修复 → 重编译，2 轮 & 第 11 章
peer-review & 同行评审（CONSORT / STROBE 规范） & 第 11 章
cn-academic-humanizer & 中文学术去 AI 痕迹（适配 CSSCI / 光明日报 / 西南金融） & 第 11 章
paper-slides & 会议演讲（beamer → PDF + PPTX + 讲稿） & 课外
paper-poster & 会议海报（A0/A1 PDF + PPTX + SVG） & 课外
nsfc-bib-manager & BibTeX 新增核验（防幻觉引用） & 第 3 章\

E.2.4 期刊定位组（Awesome-Journal-Skills，已装 25 个）★★★

• 中文 CSSCI 顶刊（12）：经济研究 er-* / 管理世界 mw-* / 中国社会科学 / 中国工业经济 / 世界经济 / 财贸经济 cte-* / 金融研究 jfr-* / 中国农村经济 cre-* / 经济学（季刊）ceq-* / 管理科学学报 jmsc-* / 数量经济技术经济研究 jqte-* / 财经研究 jfe-*

• 英文 SSCI（4）：JF / RFS / JBF / JCF

• 农经环境（9）：Agricultural Systems / Agriculture Environment / Field Crops Research / Global Change Biology / Global Environmental Change / JEEM 等

• 总计 208 个期刊子 skill（第 11 章 §11.6 投稿流程配套）

E.2.5 其他

表 E.9 其他。

L3|L3|L3

Skill & 用途 & 对应章节
questionnaire-design & 经济学 / 农村发展学术问卷设计 & 第 7 章
stata & Stata 回归分析 SOP（.do 文件 / 计量 / 图形 / 包） & 第 1 / 8 / 9 章
gzh-design-skill & 公众号排版（Markdown → 可粘贴 HTML） & 课外
notes / knowledge-base & IMA 笔记 / 知识库 & 课外\

E.3 CLI 工具（执行层）

E.3.1 Zotero 全家桶（第 3 章主轴工具）★★★

表 E.10 Zotero 全家桶（第 3 章主轴工具）。

L3|L3|L3

组件 & 版本 & 实测状态
Zotero 桌面端 & 10.0.1 & ✅ /Applications/Zotero.app；数据目录 ~/Zotero
cli-anything-zotero & Agent-native CLI & ✅ ；SQLite backend 离线实测通过
CLI Bridge 插件 & cli-bridge@cli-anything.dev & ✅ 已装（manifest 修补 strict_maxVersion 10.\）
9 款插件 & — & jasminum（中文）/ better-bibtex / zotero-pdf-translate / zotmoov / zotero-reference / better-notes / attanger / zotero-style / format-metadata\

对学生最有价值的 8 项命令能力（详见第 3 章 §3.2-3.5）：

表 E.11 命令 / 用途速查表。

L3|L3|L3

# & 命令 & 用途
1 & cli-anything-zotero add doi/arxiv/bibtex/url --fetch-pdf & 五种入库 + 自动抓全文
2 & cli-anything-zotero collection fetch-pdfs & 整分类批量补 PDF（OA 级联）
3 & cli-anything-zotero item bibliography --style "GB/T 7714" & 生成国标引用
4 & cli-anything-zotero item semantic-search & 本地语义搜索（不联网，免费）
5 & cli-anything-zotero item search-fulltext & PDF 全文检索
6 & cli-anything-zotero item duplicates + merge & 查重去重
7 & cli-anything-zotero docx cite & Word 引用管线（{{zotero:ITEMKEY}} → 真实引用字段）
8 & cli-anything-zotero export bib --format bibtex & BibTeX 导出（配 better-bibtex 供 LaTeX/Quarto 用）\

E.3.2 其他 CLI 工具

表 E.12 其他 CLI 工具。

L4|L4|L4|L4

CLI 工具 & 用途 & 安装 & 对应章节
officecli & Word / Excel / PPT 操作（学术稿优先） & pip install officecli & 第 1 / 8 / 11 章
cli-anything & 自定义 CLI 构建（任意网站 → CLI） & pip install cli-anything & 第 1 章
edge-tts & 语音合成（公众号 / 视频配音） & pip install edge-tts & 课外
pandoc & 文档格式互转（.md ↔ .docx ↔ .pdf） & brew install pandoc / winget install pandoc & 第 11 章
quarto & .ipynb → PDF / HTML / Word 多格式发布 & brew install --cask quarto & 课外
uv & Python 包管理器（极快） & brew install uv / pip install uv & 第 1 章
gh & GitHub CLI（PR / Issue / Release） & brew install gh / winget install gh & 第 1 章
git & 版本控制 & 系统自带或 brew install git & 第 1 章\

---

E.4 PDF 获取避坑五招（第 3 章 §3.6 配套）

表 E.13 PDF 获取避坑五招（第 3 章 §3.6 配套）。

L4|L4|L4|L4

招 & 场景 & 做法 & 成功率
1 & NBER 工作论文 & curl 直链 & 100%
2 & Zenodo & 原生 curl/8 UA（浏览器 UA 被 403 “unusual traffic”）；API zenodo.org/api/``records?q={query} & 高
3 & 找 OA 副本 & OpenAlex api.openalex.org/``works/doi:{doi} 的 locations 字段（机构库 / PMC / figshare） & 高
4 & MDPI（Akamai 拦截） & Wayback Machine → mdpi-res.com CDN 直链 & 稳定
5 & scansci-pdf & 单条 download 比 batch 稳定（batch 有 MCP 适配陷阱：identifiers 被包装为 {"item": [...]}） & 中高\

E.4.1 无法绕过黑名单（只能机构 WebVPN 或图书馆）

Wiley（Cloudflare）/ OUP（cookie 挑战）/ Emerald（JS 挑战）/ Preprints.org（Akamai）/ SSRN（SSL 拒绝 + 登录墙）

E.4.2 必验命令（每次下载后跑一遍）

head -c 4 file.pdf   # 必须返回 %PDF；4.8KB HTML 假文件是典型反爬产物

---

E.5 黄金组合示例（MCP × Skill × CLI）

表 E.14 黄金组合示例（MCP × Skill × CLI）。

L5|L5|L5|L5|L5

研究任务 & MCP（连接层） & Skill（知识层） & CLI（执行层） & 流程
银行面板回归 & stata-mcp & stata & officecli & 拉数据 → 跑回归 → 输出 Word 三线表
文献综述 & nber-cli-mcp + scansci-pdf + zenodo-mcp & litreview / systematic-literature-review & cli-anything-zotero & 检索 → 下载 → 入库 Zotero → 综述
问卷分析 & — & questionnaire-design & officecli & 设计问卷 → 信效度 → 输出 docx
数据可视化 & tushareMcp / faostat-mcp-server & — & mcp-server-chart + officecli & 拉数据 → 画图 → 嵌入 Word
国际数据对比 & imf-mcp-server + oecd-mcp + faostat-mcp-server & — & officecli & 三剑客查询 → 横向对比表
论文写作（投稿版） & officecli-mcp & academic-paper-writer + cn-academic-humanizer + paper-writing & officecli + pandoc & 起草 → 去 AI 腔 → 期刊风格校准 → 投稿版 docx
案例研究编码 & — & academic-paper-analysis & cli-anything-zotero（访谈稿入库） & 访谈文本 → 三段编码 → 案例矩阵
期刊投稿定位 & — & Awesome-Journal-Skills（如 mw-workflow / cre-workflow / jfr-workflow） & officecli & 选刊 → 风格对标 → 投稿信\

---

E.6 工具选型决策树

你的任务是什么？
+-- 拉数据 -> MCP（faostat/imf/oecd/tushare/nber/zenodo）
+-- 跑分析 -> MCP（stata-mcp）+ Skill（stata）
+-- 写文档 -> CLI（officecli）+ Skill（academic-paper-writer / cn-academic-humanizer）
+-- 检索文献 -> MCP（nber-cli-mcp / scansci-pdf / zenodo-mcp）+ Skill（litreview / find-literature）
+-- 管理引用 -> CLI（cli-anything-zotero）+ Zotero 桌面端
+-- 设计问卷 -> Skill（questionnaire-design）
+-- 投稿选刊 -> Skill（Awesome-Journal-Skills 期刊子 skill）
+-- 排版发布 -> CLI（pandoc / quarto）+ Skill（gzh-design-skill 公众号 / paper-slides 演讲）

[提示] 三层组合原则：单一任务通常 MCP（拉数据/远程执行）+ Skill（领域 SOP）+ CLI（本地操作）三者协同；任一层缺位会让流程断链。

---

English Edition

Appendix E · Quick-Reference Matrix of MCP / Skill / CLI Tools

---

E.1 MCP servers (connection layer)

E.1.1 Literature search / retrieval (core of Chapter 3) ★★★

Table E.1 Literature search / retrieval quick reference.

L4|L4|L4|L4

MCP name & Number of tools & Core tools & Related chapter
nber-cli-mcp & 3 & search_papers / get_paper_info / download_paper (date filters and paging) & Ch. 3
scansci-pdf & 23 & / / (push PDFs straight into Zotero) / / / / tor support & Ch. 3
zenodo-mcp & 5 & search_records / get_metadata / get_citation (citation format directly) / detect_data_type / list_files & Ch. 3 / 6\

E.1.2 International-organisation data (Chapter 6 data-source panorama) ★★

Table E.2 International-organisation data.

L4|L4|L4|L4

MCP name & Number of tools & Core tools & Use case
imf-mcp-server & 5 & imf_list_databases / imf_get_database / imf_query_dataset / imf_dataframe_query & International macro finance (IFS / WEO)
oecd-mcp & 9 & list_dataflows / search_dataflows / query_data / search_indicators / get_dataflow_url & Advanced economies + agricultural policy databases
faostat-mcp-server & 6 & / / / & FAO’s 69 agricultural domains (core of agricultural economics)\

[Tip] Selection mnemonic: for agricultural questions go to FAO, for international comparison go to IMF, for advanced economies go to OECD.

E.1.3 Econometrics and writing (Chapters 1 / 8 / 9 / 11) ★★

Table E.3 Econometrics and writing quick reference.

L4|L4|L4|L4

MCP name & Number of tools & Core tools & Related chapter
stata-mcp & 3 & stata_session / stata_run_file / stata_run_selection (official port already configured) & Ch. 1 / 8 / 9
officecli-mcp & 1 & officecli (all-round Word / PPT / Excel editing; run load_skill word academic-paper first) & Ch. 1 / 8 / 11
mcp-server-chart & 27 & 27 chart types: scatter / Sankey / mind map / Gantt / radar / Venn / word cloud / funnel, etc. & Ch. 8
sequential-thinking & 1 & sequentialthinking (structured chain of thought, stepwise argumentation) & Ch. 4\

E.1.4 Financial data (optional, brief mention)

Table E.4 Financial data (optional, brief mention).

L3|L3|L3

MCP name & Number of tools & Coverage
tushareMcp & 254 & A-shares / funds / futures / options / bonds / macro
china-stock-mcp & 30 & A-shares + sectors + financials + research reports
hexin-ifind-ds-* series & 7 sub-services, 35+ tools & stock / bond / fund / edb / global-stock / news / index\

E.1.5 Web interaction (Chapter 3 extension)

Table E.5 Web interaction (Chapter 3 extension).

L3|L3|L3

MCP name & Number of tools & Purpose
browser-use & 16 & Interaction with publisher sites and database platforms
chrome-devtools & 29 & Chrome debugging, take_screenshot, evaluate_script DOM audits\

△  CNKI boundary: this handbook does not cover cnki-mcp (locked in handbook v1.4); search Chinese literature on the CNKI website yourself and capture metadata with the Zotero jasminum plugin (Chapter 3, §3.5).

---

E.2 Skills (knowledge layer)

E.2.1 Literature search / review group (10 skills) ★★★

Table E.6 Literature search / review group quick reference.

L3|L3|L3

Skill & Description highlights & Related chapter
arxiv & arXiv search / download / abstract, saved to the local paper library & Ch. 3
litreview & Consensus search + PICO/SPIDER frameworks + Word research guide (“launching pad”) & Ch. 3
find-literature & “find papers / what does the literature say / cite papers” & Ch. 3
paper-finder-research & Allen AI PaperFinder intelligent literature review + automatic download and saving & Ch. 3
research-lit & Paper search / analysis / related work / key ideas & Ch. 3
literature-toolkit & Academic literature CLI: search, parse, and analyze & Ch. 3
lit-review-assistant & Economics literature review assistant (search, summarize, synthesize) & Ch. 3
systematic-literature-review & Multi-source search → deduplication → AI scoring → paper selection → expert writing, PDF + Word, multilingual & Ch. 3
academic-literature-guide & Four-level progressive reading: intuition → concepts → techniques → critique & Ch. 3
academic-paper-analysis & Deep PDF analysis into structured Markdown, including writing and argumentation analysis & Ch. 3 / 10\

E.2.2 Full research pipeline group (7 skills) ★★

Table E.7 Full research pipeline group.

L3|L3|L3

Skill & Description & Related chapter
academic-researcher & Systematic academic research from topic to literature synthesis & Ch. 2 / 3
research-ideation & Generate research questions from economic phenomena & Ch. 2
research-refine & Direction refinement (iterative GPT review, problem-anchored) & Ch. 4
research-review & Deep critical GPT review (via Codex MCP) & Ch. 4 / 11
research-pipeline & Workflow 1 + 2 end-to-end orchestration & Extracurricular
Academic-Research-Agent-Skill & 18-command framework: scope → reality-gate → novelty-review → claim verification & Extracurricular (high-star GitHub)
Awesome-Vibe-Research & Methodology document for the whole research pipeline (8 stages) & Extracurricular (high-star GitHub)\

E.2.3 Paper writing group (12 skills) ★★

Table E.8 Paper writing group.

L3|L3|L3

Skill & Description & Related chapter
academic-paper-writer & Draft economics papers with proper structure and academic style & Ch. 11
paper-plan & Paper outline (review conclusions → outline) & Ch. 11
paper-write & Chapter-by-chapter LaTeX drafting & Ch. 11
paper-figure & Paper figure generation & Ch. 8 / 11
paper-compile & LaTeX compilation + fixes + verification & Ch. 11
paper-writing & Workflow 3 end-to-end orchestration (plan → figure → write → compile → polish loop) & Ch. 11
auto-paper-improvement-loop & GPT review → fixes → recompile, two rounds & Ch. 11
peer-review & Peer review (CONSORT / STROBE standards) & Ch. 11
cn-academic-humanizer & Remove AI traces from Chinese academic prose (tuned for CSSCI / Guangming Daily / Southwest Finance) & Ch. 11
paper-slides & Conference talks (beamer → PDF + PPTX + speaker notes) & Extracurricular
paper-poster & Conference posters (A0/A1 PDF + PPTX + SVG) & Extracurricular
nsfc-bib-manager & BibTeX addition verification (guards against hallucinated citations) & Ch. 3\

E.2.4 Journal-targeting group (Awesome-Journal-Skills, 25 installed) ★★★

• Top Chinese CSSCI journals (12): Economic Research er-* / Management World mw-* / Social Sciences in China / China Industrial Economics / The Journal of World Economy / Finance & Trade Economics cte-* / Journal of Financial Research jfr-* / Chinese Rural Economy cre-* / China Economic Quarterly ceq-* / Journal of Management Sciences in China jmsc-* / The Journal of Quantitative & Technological Economics jqte-* / Journal of Finance and Economics jfe-*

• English SSCI (4): JF / RFS / JBF / JCF

• Agricultural and environmental (9): Agricultural Systems / Agriculture Environment / Field Crops Research / Global Change Biology / Global Environmental Change / JEEM and others

• 208 journal sub-skills in total (supporting the submission workflow in Chapter 11, §11.6)

E.2.5 Others

Table E.9 Others.

L3|L3|L3

Skill & Purpose & Related chapter
questionnaire-design & Academic questionnaire design for economics / rural development & Ch. 7
stata & Stata regression analysis SOP (.do files / econometrics / graphics / packages) & Ch. 1 / 8 / 9
gzh-design-skill & WeChat article layout (Markdown → paste-ready HTML) & Extracurricular
notes / knowledge-base & IMA notes / knowledge base & Extracurricular\

E.3 CLI tools (execution layer)

E.3.1 The Zotero suite (main tool of Chapter 3) ★★★

Table E.10 The Zotero suite.

L3|L3|L3

Component & Version & Verified status
Zotero desktop & 10.0.1 & ✅ /Applications/Zotero.app; data directory ~/Zotero
cli-anything-zotero & Agent-native CLI & ✅ ; SQLite backend verified offline
CLI Bridge plugin & cli-bridge@cli-anything.dev & ✅ installed (manifest patched for strict_maxVersion 10.\)
9 plugins & — & jasminum (Chinese) / better-bibtex / zotero-pdf-translate / zotmoov / zotero-reference / better-notes / attanger / zotero-style / format-metadata\

The eight command capabilities most valuable to students (see Chapter 3, §3.2–3.5):

Table E.11 Commands / purposes quick reference.

L3|L3|L3

# & Command & Purpose
1 & cli-anything-zotero add doi/arxiv/bibtex/url --fetch-pdf & Five ways to ingest + automatic full-text retrieval
2 & cli-anything-zotero collection fetch-pdfs & Batch PDF top-up for a whole collection (OA cascade)
3 & cli-anything-zotero item bibliography --style "GB/T 7714" & Generate GB/T 7714 citations
4 & cli-anything-zotero item semantic-search & Local semantic search (offline, free)
5 & cli-anything-zotero item search-fulltext & PDF full-text search
6 & cli-anything-zotero item duplicates + merge & Duplicate detection and merging
7 & cli-anything-zotero docx cite & Word citation pipeline ({{zotero:ITEMKEY}} → real citation fields)
8 & cli-anything-zotero export bib --format bibtex & BibTeX export (with better-bibtex for LaTeX/Quarto)\

E.3.2 Other CLI tools

Table E.12 Other CLI tools.

L4|L4|L4|L4

CLI tool & Purpose & Installation & Related chapter
officecli & Word / Excel / PPT operations (academic manuscripts first) & pip install officecli & Ch. 1 / 8 / 11
cli-anything & Build custom CLIs (any website → CLI) & pip install cli-anything & Ch. 1
edge-tts & Speech synthesis (WeChat articles / video voice-over) & pip install edge-tts & Extracurricular
pandoc & Document format conversion (.md ↔ .docx ↔ .pdf) & brew install pandoc / winget install pandoc & Ch. 11
quarto & .ipynb → PDF / HTML / Word multi-format publishing & brew install --cask quarto & Extracurricular
uv & Python package manager (extremely fast) & brew install uv / pip install uv & Ch. 1
gh & GitHub CLI (PR / Issue / Release) & brew install gh / winget install gh & Ch. 1
git & Version control & Built in or brew install git & Ch. 1\

---

E.4 Five tricks for getting PDFs without pitfalls (supporting Chapter 3, 3.6)

Table E.13 Five tricks for getting PDFs.

L4|L4|L4|L4

Trick & Scenario & Method & Success rate
1 & NBER working papers & Direct curl link & 100%
2 & Zenodo & Native curl/8 UA (a browser UA gets 403 “unusual traffic”); API zenodo.org/api/``records?q={query} & High
3 & Finding OA copies & The locations field of OpenAlex api.openalex.org/``works/doi:{doi} (institutional repositories / PMC / figshare) & High
4 & MDPI (Akamai blocking) & Wayback Machine → the mdpi-res.com CDN direct link & Stable
5 & scansci-pdf & Single download is more stable than batch (batch has an MCP adaptation trap: identifiers are wrapped as {"item": [...]}) & Moderately high\

E.4.1 Blacklists that cannot be bypassed (institutional WebVPN or the library only)

Wiley (Cloudflare) / OUP (cookie challenge) / Emerald (JS challenge) / Preprints.org (Akamai) / SSRN (SSL rejection + login wall)

E.4.2 Mandatory check (run after every download)

head -c 4 file.pdf   # must return %PDF; a 4.8 KB HTML fake file is the typical anti-scraping product

---

E.5 Golden combinations (MCP × Skill × CLI)

Table E.14 Golden combination examples (MCP × Skill × CLI).

L5|L5|L5|L5|L5

Research task & MCP (connection) & Skill (knowledge) & CLI (execution) & Workflow
Bank panel regression & stata-mcp & stata & officecli & Pull data → run regression → output a three-line Word table
Literature review & nber-cli-mcp + scansci-pdf + zenodo-mcp & litreview / systematic-literature-review & cli-anything-zotero & Search → download → import into Zotero → review
Questionnaire analysis & — & questionnaire-design & officecli & Design questionnaire → reliability and validity → output docx
Data visualisation & tushareMcp / faostat-mcp-server & — & mcp-server-chart + officecli & Pull data → chart → embed in Word
International data comparison & imf-mcp-server + oecd-mcp + faostat-mcp-server & — & officecli & Query all three → side-by-side comparison table
Paper writing (submission version) & officecli-mcp & academic-paper-writer + cn-academic-humanizer + paper-writing & officecli + pandoc & Draft → de-AI the prose → calibrate to journal style → submission docx
Case-study coding & — & academic-paper-analysis & cli-anything-zotero (import interview transcripts) & Interview text → three-stage coding → case matrix
Journal targeting & — & Awesome-Journal-Skills (e.g. mw-workflow / cre-workflow / jfr-workflow) & officecli & Pick a journal → benchmark style → cover letter\

---

E.6 Tool-selection decision tree

What is your task?
+-- Pull data     -> MCP (faostat/imf/oecd/tushare/nber/zenodo)
+-- Run analysis  -> MCP (stata-mcp) + Skill (stata)
+-- Write docs    -> CLI (officecli) + Skill (academic-paper-writer / cn-academic-humanizer)
+-- Search papers -> MCP (nber-cli-mcp / scansci-pdf / zenodo-mcp) + Skill (litreview / find-literature)
+-- Manage citations -> CLI (cli-anything-zotero) + Zotero desktop
+-- Design survey -> Skill (questionnaire-design)
+-- Target a journal -> Skill (Awesome-Journal-Skills journal sub-skills)
+-- Typeset and publish -> CLI (pandoc / quarto) + Skill (gzh-design-skill for WeChat / paper-slides for talks)

[Tip] Three-layer combination principle: a single task usually needs MCP (pulling data / remote execution) + Skill (domain SOP) + CLI (local operations) working together; if any layer is missing, the workflow breaks.