Paper-WorkFlow skill
An orchestrator skill that takes empirical social science work from idea or data to a reproducible manuscript with quality gates.
- 50 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill runs analysis in Python, Stata or R and writes workspace files, and sub-agents work on your data.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/paper-workflowDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill runs analysis in Python, Stata or R and writes workspace files, and sub-agents work on your data.
This is third-party code. Review the repository files before installing.
What it does
Paper-WorkFlow runs an empirical study through ten stages, 0 to 9, with two hard quality gates and a single workspace. The skill does not rewrite every capability and instead routes existing skills, analysis backends and subagents. You can enter from anywhere: an idea, a proposal, data, finished regressions, a `main.tex` file or reviewer comments. Stages cover topic choice, data, identification design, estimation, tables and figures, a draft, structural and language polish, a simulated review and a submission package. Analysis runs on Python, Stata or R with fixest. Results land in a workspace with state, and `pw check`, `pw exit N` and `pw final` verify readiness. The skill promises neither significant results nor acceptance, and failure is also recorded.
Who it is for. Economics and social science researchers running empirical work who want to go from data to manuscript with a decision log and gates.
Good fit when
- You only have a topic and need to check whether data exists and the design is feasible
- Data is in a panel file, the design is DiD and a baseline estimate with robustness checks is needed
- You have `main.tex` and want to continue from the polish stage
- Tables and figures are needed with reproducible code and a quality report
- Results must be rebuilt into a DOCX manuscript noting that the analysis is retrospective
Not a fit when
- You need a guarantee of significant results or publication, which the skill does not give
- The work is not empirical social science: the orchestrator is built for such studies
- You need the full set without sub-skills: a full run needs the parent collection of 47 skills
Example request
/paper-workflow Data is in `./panel.csv` with a DiD design. First audit the sample and variables, then run a baseline estimate and robustness checks.Limitations
The orchestrator in its standalone repository without sub-skills does not run fully: for the complete chain cloning the parent collection Auto-Empirical-Research-Skills with submodules is recommended. End to end means stage coverage, not a guarantee of data availability, significance, identification checks passing or acceptance. Running Stata requires a license, and R and Python are needed for their backends. The language-stage the author calls removing AI traces remains editorial polish and no detector evasion. Checks run through `pw`, and DOCX needs pandoc. The license is MIT.
How to disable. Delete the skill folder from `~/.claude/skills/` or the project `.claude/skills/`. Research workspaces with data and results are yours and are deleted separately.
FAQ
How do I install Paper-WorkFlow?
Cloning the parent collection is better: `git clone --recurse-submodules https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills.git`. The orchestrator alone installs with `git clone https://github.com/brycewang-stanford/Paper-WorkFlow.git`.
Where can I start?
From an idea, a proposal, clean data, finished regressions, a `main.tex` file or reviewer comments. The skill joins at the matching stage.
What do `pw check`, `pw exit` and `pw final` do?
`pw check` shows the current state, `pw exit N` checks stage completion and `pw final` requires all submission obligations to be done.
Which analysis backends exist?
Python, Stata and R with fixest and modelsummary. You choose one in the request, for example "use Stata".
Does the skill promise publication?
No. The skill guarantees no significant results, identification checks passing or acceptance, and failures are recorded in the workspace.
Discussion
No questions or reviews yet.
Related
SkillsThe paper-spine skill takes a topic, materials and data through a scientific paper: literature, text, figures, review and Word, LaTeX and PDF files
Skill with 30 prompt templates for research papers in English and Chinese, from literature review to grant proposals
Eight skills draft a CS, AI or ML paper section by section and halt any section that lacks supporting evidence.
A skill that stress-tests a research paper before submission: mocks a review, gives a verdict, helps revise and re-verify
A Codex skill set for research papers: from research assets to a LaTeX draft with evidence checks and a review simulation.
42 skills for the whole research paper path: from discovery and drafting to citation checks, submission, rebuttal and publication
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Paper-WorkFlow (this card) | no | 50 | |
| Auto-Empirical Research Skills | no | 5,007 | the parent collection of 47 skills; Paper-WorkFlow is its orchestrator as a separate repository with stages and gates |
| Nature-Paper-Skills | no | 671 | skills for paper structure and prose; this covers the whole chain from data to submission |
README in short
The README in a 30-second format explains that this is an orchestrator, lists stages, entry points and backends, shows sample prompts, the output materials and installation. Separate files describe the strict gates.
README badge
Original SKILL.md
---
name: paper-workflow
description: >
将社科实证研究从 idea、proposal、数据或已有结果推进到可复现分析和完整论文 DOCX。
支持研究设计、数据审计、估计、表图、写作与修订，以及按需准备投稿包；按证据完成阶段，
不承诺显著结果或录用。触发：/paper-workflow、从想法到论文、从数据到 docx、推进已有实证稿件。
allowed-tools: Skill, Agent, Read, Write, Edit, Glob, Grep, Bash, AskUserQuestion, WebSearch, WebFetch, NotebookEdit
---
# Paper-WorkFlow — 经管 / 社科实证论文全流程总编排器
## Overview
本 skill 按 **Stage 0–9 + Method Gate + Draft Quality Gate** 组织社科实证研究。
优先调用可用子 skill；缺少外部技能库或 Agent 时，按本仓库 references 由主代理直接执行。
方法、数据权限与统计软件仍须真实可用；不能以补齐模板替代研究证据。
`workflow_state.json`、stage passport 和 handoff 保存状态与文件指针。调用映射见
[`references/skill-map.md`](references/skill-map.md)，当前阶段操作见
[`references/stage-playbook.md`](references/stage-playbook.md)，独立安装路径见
[`references/runtime-fallbacks.md`](references/runtime-fallbacks.md)。只加载当前阶段需要的部分。
---
## 这条流水线（固定主线，可按入口跳入）
| Stage | 阶段 | 主要调用的 skill（位于 `67-econfin-workflow-toolkit/`，除非另注） | 产出落盘 |
|---|---|---|---|
| **0** | Intake & Setup | *(编排器本体)* | 工作区、`workflow_state.json`、入口路由 |
| **1** | 选题与设计 | `econfin-idea-finder` → `novelty-check` → `significance-search` → `journal-digest` → `econfin-proposal` | `01_proposal/proposal.md` |
| **1L** | 文献基座（先于查新打分） | 多路并行检索：`59-openalex` / `36` / `52` + Zotero MCP | `01_proposal/lit/corpus.md` + `lit_matrix.md` |
| **2** | 数据 | `data-fetcher` → `data-cleaning` + sample/estimand audit + 测量效度 | `02_data/clean.parquet` + `codebook.md` + `sample_audit.md` + `measurement_audit.md` |
| **2.5** | 设计锁定（先于第一个估计） | *(编排器本体，人类决策点)* | `00_meta/preregistration.md` + `design_lock` 状态 |
| **3** | 计量识别、估计与方法闸门 | **分析后端路由**：默认 Python/StatsPAI（MCP 优先，调用链见 `statspai-analysis.md`），也可切 Stata（`.do`）或 R（fixest/Quarto）；再按设计配 `did-analysis` / `iv-estimation` / `rdd-analysis` / `synthetic-control` / `panel-data` / `ml-causal` 等（全表见 skill-map）+ empirical audit + methods pack + design gate cards | `03_analysis/` 代码 + `design_register.md` + `method_gate.md` + `evidence_ledger.md` |
| **4** | 表与图 | 按同一后端生成出版级表图：Python/StatsPAI `regtable`/`paper_tables`/`collect`，Stata `esttab`/`outreg2`/`collect`，R `modelsummary`/`etable`/Quarto；均需三线表 Word/Excel/LaTeX 同出 + PDF/PNG 图 | `04_results/*.{tex,docx,xlsx}` + `*.pdf/png` |
| **5** | 写作初稿 | `paper-writer` | `05_draft/main.md`（缺省；`latex` 轨为 `main.tex`）+ `ref.bib`；表图一律 include 引用 |
| **6** | **结构层**打磨（段落及以上） | `paper-pipeline`（内部跑 polish→self-revise→style→polish→reference-verify 全量基线） | 打磨后的 `main.{tex,md}` |
| **7** | **语言层**去 AI 味（句子及以下） | `de-aigc-skills`(48 中英双语六步闭环) → `readability`/`fix-chinese` 收尾 | 降味稿 + 审计表 + 数字零漂移 + `00_meta/ai_use_disclosure.md` 补上本阶段台账行 |
| **8** | 模拟评审与修订 | `referee-report` → `paper-referee-revise`（或 `paper-self-revise`） | 修订稿 + response letter |
| **9** | 选刊与投稿 + **全文交付** | `paper-submission` → 可选 AJS 适配器 → `reference-verify`（终审）→ `assemble_manuscript_docx.py` | 期刊清单 + 专属预检 + cover letter + AI 使用声明 + **`09_submission/main.docx` 全文定稿** |
| **—** | 复盘与交付 | *(编排器本体)* | `FINAL_REPORT.md` + 打包交付物 |
> 完整阶段细节、每阶段的 plan→execute→review→revise 微循环、subagent 派发模板，全部在
> [`references/stage-playbook.md`](references/stage-playbook.md)。**主代理在进入某阶段时才去读
> 对应章节**。