[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82583":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":24,"hasPages":22,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":43,"readmeContent":44,"aiSummary":45,"trendingCount":15,"starSnapshotCount":15,"syncStatus":46,"lastSyncTime":47,"discoverSource":48},82583,"Auto-Empirical-Research-Skills","brycewang-stanford\u002FAuto-Empirical-Research-Skills","brycewang-stanford","🔬 A curated collection of 23,000+ agent skills for empirical research across 8 social science disciplines. | 精选 23,000+ AI Agent 技能库，覆盖8大社会科学学科的实证研究。CoPaper.AI 20分钟完成一篇可复现的规范实证论文，并支持用户上传 Skills。-- Maintained by CoPaper.AI from Stanford REAP.","https:\u002F\u002Fcopaper.ai",null,"Stata",1799,266,7,0,89,356,431,267,20.28,"Other",false,"main",true,[26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"academic-research","agent-skills","ai-agent","awesome-list","communication","copaper","economics","education","empirical-research","international-relations","political-science","psychology","public-administration","reproducible-research","skills-library","social-science","sociology","2026-06-12 02:04:26","# Auto-Empirical Research Skills (AERS, 23K+ Skills)\n\n> [!NOTE]\n> **Repository renamed → now \"Auto\".** This project started life as *Awesome Agent Skills for Empirical Research* and has been renamed to **Auto-Empirical-Research-Skills (AERS)**. The new name reflects the core idea: not just a *collection* of skills, but an agent that **automatically runs the full empirical-research pipeline end to end** — from raw data cleaning → identification & estimation → robustness checks → tables, figures, and a submission-ready draft — with minimal human hand-holding.\n>\n> GitHub automatically redirects the old URL, but please update your bookmarks and local remote:\n>\n> ```bash\n> git remote set-url origin https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills.git\n> ```\n\n\u003Cdiv align=\"center\">\n\n**🌐 Language \u002F 语言: English | [中文](README-zh.md)**\n\n\u003Cbr\u002F>\n\n  \u003Cimg src=\"images\u002Faers-readme-cover-en.png\" alt=\"Auto-Empirical Research Skills cover\" width=\"100%\" \u002F>\n\n  \u003Cbr\u002F>\n\n  \u003Ctable>\n    \u003Ctr>\n      \u003Ctd align=\"center\">\n        \u003Ca href=\"https:\u002F\u002Fcopaper.ai\">\u003Cimg src=\"images\u002Fcopaper-logo.png\" alt=\"CoPaper.AI\" width=\"260\" \u002F>\u003C\u002Fa>\n      \u003C\u002Ftd>\n      \u003Ctd width=\"60\">\u003C\u002Ftd>\n      \u003Ctd align=\"center\">\n        \u003Cimg src=\"images\u002Fstanford-reap-logo.png\" alt=\"Stanford REAP - Center on China's Economy & Institutions\" width=\"380\" \u002F>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftable>\n\n  \u003Cbr\u002F>\n\n  \u003Cstrong>Stanford REAP × CoPaper.AI\u003C\u002Fstrong> · An academic-industrial AI toolkit for empirical research\u003Cbr\u002F>\n  \u003Csub>Crafted by Stanford's empirical methodology team — covering the full pipeline from data cleaning to top-journal submission\u003C\u002Fsub>\n\n  \u003Cbr\u002F>\n\u003C\u002Fdiv>\n\n[![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re)\n[![GitHub stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills?style=social)](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills)\n[![License: CC BY-SA 4.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY--SA%204.0-lightgrey.svg)](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-sa\u002F4.0\u002F)\n[![PRs Welcome](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPRs-welcome-brightgreen.svg)](CONTRIBUTING.md)\n[![Maintained by CoPaper.AI from Stanford REAP](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FMaintained%20by-CoPaper.AI%20from%20Stanford%20REAP-blue)](https:\u002F\u002Fcopaper.ai)\n[![Powered by StatsPAI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPowered%20by-StatsPAI-orange)](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI)\n[![Security Scanned](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSecurity-52%2F52%20CLEAN-brightgreen)](SECURITY-SCAN-REPORT.md)\n[![Files Audited](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Ffiles%20audited-2%2C940%2B-blue)](SECURITY-SCAN-REPORT.md)\n[![Audit Phases](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Faudit-6%20phases%20%2F%2013%20risk%20categories-blueviolet)](SECURITY-SCAN-REPORT.md)\n[![Hooks Audited](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fhook%20scripts-40%2B%20audited-blue)](SECURITY-SCAN-REPORT.md)\n[![Zero Threats](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fthreats%20found-0-brightgreen)](SECURITY-SCAN-REPORT.md)\n[![Validate catalog](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills\u002Factions\u002Fworkflows\u002Fvalidate-catalog.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills\u002Factions\u002Fworkflows\u002Fvalidate-catalog.yml)\n[![OpenSSF Scorecard](https:\u002F\u002Fapi.scorecard.dev\u002Fprojects\u002Fgithub.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills\u002Fbadge)](https:\u002F\u002Fscorecard.dev\u002Fviewer\u002F?uri=github.com\u002Fbrycewang-stanford\u002FAuto-Empirical-Research-Skills)\n\n**The Definitive Collection of AI Agent Skills for Empirical Research — 119 GitHub Repos \u002F 23,000+ Skills**\n\n> A curated, opinionated list of **119 GitHub repositories** and **23,000+ AI Agent Skills** for empirical research in economics, political science, sociology, psychology, public health, education, management, finance, and public policy — organized by research workflow, from topic selection to journal submission.\n\nIn 2026, the way we do empirical research is being redefined.\n\n[**CoPaper.AI**](https:\u002F\u002Fcopaper.ai) — **an empirical research AI assistant incubated by researchers at [Stanford REAP \u002F SCCEI (Stanford Center on China's Economy and Institutions)](https:\u002F\u002Fsccei.fsi.stanford.edu\u002Freap)** — can **complete a publication-quality empirical paper in 20 minutes**: from data import, descriptive statistics, causal inference models, and robustness checks to formatted result tables, all in one go. The secret isn't a more powerful model — it's **Skills**: encoding senior researchers' methodological expertise into structured workflows, so the AI knows \"what a complete DID analysis should include\" instead of waiting for you to remind it step by step.\n\nThis repository is the **Agent Skills landscape** we compiled while building CoPaper.AI. We organized hundreds of Skills repos and tens of thousands of Skills scattered across GitHub, communities, and academia by research workflow stages, so you can pick what you need.\n\n**🎓 Three Layers of Trust · Why It's Us Building This**\n\n| Layer | Anchor | Lever |\n|---|---|---|\n| 🏛️ **Academic lineage** | **Stanford REAP \u002F SCCEI** — Stanford Center on China's Economy and Institutions | A research center with a sustained publication record in empirical economics methodology and a deep tradition in applied causal inference |\n| 🔧 **Engineering delivery** | **[CoPaper.AI](https:\u002F\u002Fcopaper.ai)** empirical research AI assistant | Ships with **20 econometric methodology Skills** (DID\u002FIV\u002FRDD\u002FPSM\u002FDML, etc.), Supervisor + 4 sub-agent multi-agent architecture, one-sentence triggers, automatic result output |\n| ⚙️ **Open-source engine** | **[StatsPAI](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI)** — **the causal-inference engine that powers CoPaper.AI** | **900+ functions · one `import statspai as sp` · JOSS in submission · MIT-licensed**. Every DID\u002FIV\u002FRD\u002FSCM estimate CoPaper.AI produces is driven by StatsPAI; this Skills collection is itself part of the StatsPAI ecosystem |\n\n> 🔒 **Use with confidence**: every one of the 52 Skills \u002F 2,940+ files in this repo passed our [systematic security audit](SECURITY-SCAN-REPORT.md) — **52\u002F52 CLEAN, zero FLAGGED**, zero exfiltration, zero reverse shells, zero prompt injection.\n>\n> 💡 **Want it out of the box?** Skip the Skills assembly — try [**→ copaper.ai**](https:\u002F\u002Fcopaper.ai) and let the Stanford methodology team run the empirical pipeline end-to-end for you.\n\n---\n\n## Start Here\n\n- Search the local index: [`docs\u002Fsearch.html`](docs\u002Fsearch.html)\n- Browse the generated local catalog: [`docs\u002FSKILL_CATALOG.md`](docs\u002FSKILL_CATALOG.md)\n- Copy a ready-to-run empirical workflow: [`docs\u002FGOLDEN_WORKFLOWS.md`](docs\u002FGOLDEN_WORKFLOWS.md)\n- See flagship demos: [`docs\u002Fdemos\u002F`](docs\u002Fdemos\u002F)\n- Run flagship regression prompts: [`docs\u002FEVALS.md`](docs\u002FEVALS.md)\n- Install or copy skills into an agent runtime: [`docs\u002FINSTALL.md`](docs\u002FINSTALL.md)\n- Use the machine-readable index: [`catalog\u002Fskills.json`](catalog\u002Fskills.json)\n- Coordinate parallel agent work: [`docs\u002FAGENT_COORDINATION.md`](docs\u002FAGENT_COORDINATION.md)\n- Check provenance and license risk: [`docs\u002FLICENSE_AUDIT.md`](docs\u002FLICENSE_AUDIT.md)\n- Check contribution and validation rules: [`docs\u002FQUALITY_GATE.md`](docs\u002FQUALITY_GATE.md) · [`docs\u002FSKILL_SUBMISSION_GUIDE.md`](docs\u002FSKILL_SUBMISSION_GUIDE.md)\n- See the repo audit and improvement plan: [`docs\u002FREPO_AUDIT_2026-05-31.md`](docs\u002FREPO_AUDIT_2026-05-31.md) · [`docs\u002FROADMAP.md`](docs\u002FROADMAP.md)\n- Rebuild and validate locally:\n\n```bash\nmake catalog\nmake validate\n```\n\n### Pick a Workflow in 30 Seconds\n\n| Goal | Start with |\n|---|---|\n| Run a complete empirical pipeline | [`StatsPAI_skill`](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002FSKILL.md) |\n| Audit a top-5 economics identification strategy | [`aer-identification`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-identification\u002FSKILL.md) |\n| Prepare AER \u002F AEJ submission | [`aer-workflow`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-workflow\u002FSKILL.md) |\n| Build a replication package | [`aer-replication`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-replication\u002FSKILL.md) |\n| Lower Chinese academic AI-writing signal | [`chinese-de-aigc`](skills\u002F48-copaper-ai-chinese-de-aigc\u002FSKILL.md) |\n\n---\n\n## 🆕 Changelog\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-05-25: 📕 AER-skills vendored — Top-5 economics (AER \u002F AER:Insights \u002F AEJ) submission stack (9 skills, skills\u002F50) + weekly auto-sync workflow\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **📕 [skills\u002F50-brycewang-aer-skills](skills\u002F50-brycewang-aer-skills\u002F)**: This repo's sister project [brycewang-stanford\u002FAER-skills](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAER-skills) vendored in whole, with the same StatsPAI-style sync loop ([`scripts\u002Fsync-aer-skills.sh`](scripts\u002Fsync-aer-skills.sh) + [`.github\u002Fworkflows\u002Fsync-aer-skills.yml`](.github\u002Fworkflows\u002Fsync-aer-skills.yml) — Monday 06:00 UTC weekly diff, PR on drift). **Positioning**: a top-5 economics submission skill stack (AER \u002F AER:Insights \u002F AEJ family), extending the StatsPAI \u002F 00.x \"analysis\" line to the \"manuscript + submission\" line.\n  - **🧱 Nine skills covering the full submission pipeline**: [`aer-topic-selection`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-topic-selection\u002F) (AER vs Insights vs AEJ routing) → [`aer-identification`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-identification\u002F) (identification audit: modern DiD \u002F weak IV \u002F boundary RDD pitfalls) → [`aer-robustness`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-robustness\u002F) (referee-anticipating robustness matrix) → [`aer-introduction`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-introduction\u002F) (Keith Head five-paragraph intro) → [`aer-tables-figures`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-tables-figures\u002F) (AER booktabs typesetting) → [`aer-replication`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-replication\u002F) (AEA Data and Code Availability Policy package, openICPSR-ready) → [`aer-submission`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-submission\u002F) (preflight: 100-word abstract, disclosure, cover letter) → [`aer-rebuttal`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-rebuttal\u002F) (R&R letters written against the *revised* manuscript, not the old draft) → [`aer-workflow`](skills\u002F50-brycewang-aer-skills\u002Fskills\u002Faer-workflow\u002F) (orchestrator that tells you which skill to use next).\n  - **🆚 Differentiation from existing skills**: StatsPAI \u002F 00.x solve \"how to run the analysis correctly\"; AER-skills solves \"how to write the paper to top-5 acceptance threshold\" — the AER 100-word abstract \u002F AER:Insights 7000-word limit \u002F 45% desk-rejection rate \u002F AEA mandatory replication are top-5-specific constraints that generic scientific-writing skills (Nature-Paper-Skills etc.) do not cover. **Identification-first**: if your design is fragile, no prose will save it.\n  - **🔁 Vendor-sync loop**: `git clone --depth=1` upstream → `rsync -a --delete --exclude='.git'` mirror the whole tree → diff content hashes before\u002Fafter, exit 0 on no drift, exit 1 on drift to trigger `peter-evans\u002Fcreate-pull-request@v6` on `chore\u002Fsync-aer-skills` branch. **Supports manual `workflow_dispatch`** for on-demand sync.\n  - **License: MIT** — consistent with StatsPAI \u002F 00.x; commercial and academic use both allowed.\n  - **First upstream commit**: [`7e9c44d`](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAER-skills\u002Fcommit\u002F7e9c44d363c185edf27859096268b6a8256c4a2b) (2026-05-25, includes modern-aer-exemplars.md with 30+ subfield-organized papers).\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-28: 🛡️ Repository-wide security scan complete — all 52 Skills CLEAN, zero FLAGGED\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **🛡️ [SECURITY-SCAN-REPORT.md](SECURITY-SCAN-REPORT.md)**: We ran a systematic security audit across **all 52 Skills \u002F 2,940+ files** in this repository. **No malicious prompts, viruses, trojans, reverse shells, or other malicious content were found.** Bottom line: every Skill in this repo is safe to use.\n  - **🔍 Six-phase defense-in-depth methodology**: (1) automated grep across 13 risk categories (pipe-to-shell, reverse shell, credential exfil, decode-and-run, mining\u002FRAT signatures, prompt injection, etc.) → (2) 100% manual review of all 6 hook-bearing Skills and their 40+ hook scripts → (3) three parallel agents auditing SKILL.md prose, agent definitions, and reference docs separately → (4) supplemental integrity checks (hidden Unicode, encoding anomalies, ultra-long lines, HTML injection, network-related imports).\n  - **📊 Result distribution**: every \"sensitive\" hit verified as one of three legitimate categories — **defensive security rules** (deny rules, bash-safety hooks, credential detectors), **legitimate academic API calls** (arXiv \u002F CrossRef \u002F PubMed \u002F FRED \u002F World Bank \u002F OECD \u002F BLS), or **standard Claude Code workflow hooks** (scaffolding \u002F state save \u002F context monitor — **all local file operations, zero network IO**).\n  - **🔑 Key insight**: **17-DAAF is actually the strongest \"security-aware\" reference template** in this batch (14 defensive hooks + 32 deny rules + active credential scanning). Largest size ≠ highest risk.\n  - **📈 Visual infographics**: 5 zhihu-style information graphics embedded in the report ([overview](images\u002Fsecurity-scan\u002Fsecurity-scan-01-总览.png) \u002F [methodology](images\u002Fsecurity-scan\u002Fsecurity-scan-02-扫描方法.png) \u002F [threat matrix](images\u002Fsecurity-scan\u002Fsecurity-scan-03-威胁矩阵.png) \u002F [Top 5 size distribution](images\u002Fsecurity-scan\u002Fsecurity-scan-04-规模分布.png) \u002F [supplemental scan](images\u002Fsecurity-scan\u002Fsecurity-scan-05-补扫结果.png)) — readable in 3 seconds.\n  - See the [**full security scan report**](SECURITY-SCAN-REPORT.md) for details.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-24: 📗 Full Empirical Analysis Skill (R) shipped — tidyverse + fixest, 8-step Quarto-friendly loop (skills\u002F00.3)\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **📗 [Full Empirical Analysis Skill — R](skills\u002F00.3-Full-empirical-analysis-skill_R\u002F)**: Same-day fourth member of the family, vendored at [`skills\u002F00.3-Full-empirical-analysis-skill_R\u002F`](skills\u002F00.3-Full-empirical-analysis-skill_R\u002F) — **slot #0.3, the R \u002F Quarto edition**.\n  - **🧱 Modern tidyverse + fixest stack**: `dplyr` \u002F `tidyr` \u002F `haven` for data; `fixest::feols\u002Ffeglm\u002Ffepois` as the panel\u002FIV\u002FDID workhorse (one line for HD FE + multi-way cluster + IV); `did::att_gt` + `fixest::sunab` + `didimputation::did_imputation` + `synthdid` + `DIDmultiplegtDYN` + `bacondecomp` + `HonestDiD` for modern DID; `rdrobust` \u002F `rddensity` \u002F `rdmulti` \u002F `rdlocrand` for RD; `Synth` \u002F `gsynth` \u002F `tidysynth` \u002F `synthdid` for synthetic control; `MatchIt` \u002F `WeightIt` \u002F `cobalt` \u002F `ebal` for matching; `grf::causal_forest` + `DoubleML` for ML causal; `mediation::mediate` + `lavaan::sem` for mediation; `marginaleffects::avg_slopes` \u002F `plot_slopes` for post-estimation; `modelsummary` \u002F `kableExtra` \u002F `gt` \u002F `flextable` for publication tables; `ggplot2` + `iplot` + `binsreg` + `cowplot` + `patchwork` for figures; `Quarto` to render PDF\u002FHTML\u002FWord in one command.\n  - **🔁 8-step R closed loop (mirrors 00.1 \u002F 00.2)**: (1) Import & cleaning (`read_dta` + `clean_names` + `naniar::vis_miss` + `mice` + `validate` \u002F `assertr`) → (2) Variable construction (`mutate` + `across` + `DescTools::Winsorize` + `scale` + `arrange %>% group_by %>% lag\u002Flead`) → (3) Descriptives (`gtsummary::tbl_summary` + `modelsummary::datasummary_balance` + `psych::corr.test` + `corrplot` \u002F `ggcorrplot`) → (4) Diagnostics (12 classes: `shapiro.test` \u002F `tseries::jarque.bera.test` \u002F `lmtest::bptest` \u002F `dwtest` \u002F `bgtest` \u002F `car::vif` \u002F `tseries::adf.test` \u002F `kpss.test` \u002F `plm::pbgtest` \u002F `pcdtest` \u002F `phtest` \u002F `lmtest::resettest`) → (5) Estimation (12 classes: `feols` + `AER::ivreg` + `did::att_gt` + `fixest::sunab` + `didimputation` + `synthdid` + `rdrobust` + `tidysynth` + `gsynth` + `MatchIt` + `WeightIt` + `ebal` + `grf::causal_forest` + `DoubleML` + `sampleSelection::heckit` + `quantreg::rq` + `lavaan::sem`) → (6) Robustness (`modelsummary` for M1–M6 + `clubSandwich` + `fwildclusterboot::boottest` + `ri2::conduct_ri` + `bacondecomp::bacon` + `HonestDiD::createSensitivityResults` + `robomit::o_test\u002Fo_beta`) → (7) Further analysis (formula interactions + `marginaleffects::plot_slopes` + `mediation::mediate` + `medsens` + `lavaan::sem` multi-group + `grf::causal_forest` CATE + `splines::ns` dose-response) → (8) Publication output (`modelsummary` to LaTeX\u002FWord\u002FHTML\u002FMarkdown in one call + `fixest::iplot` + `marginaleffects::plot_slopes\u002Fpredictions` + `cowplot::plot_grid` + `patchwork` + `Quarto` rendering).\n  - **📚 Progressive disclosure + Quarto-native**: `SKILL.md` 893-line spine (with full `install.packages` list, project skeleton, Quarto YAML template); 8 [`references\u002FNN-*.md`](skills\u002F00.3-Full-empirical-analysis-skill_R\u002Freferences\u002F) totalling 3700+ lines. The Quarto template makes \"narrative + code + tables + figures\" render to a single self-contained report from a single `.qmd` source.\n  - **🆚 Four-skill positioning**: StatsPAI = Python one-shot DSL; 00.1 = explicit Python; 00.2 = explicit Stata; 00.3 = **R + tidyverse + Quarto**. Four parallel implementations of the same 8 steps, none replacing the others. **The Quarto-rendered reproducibility report is unique to 00.3.**\n  - **Use cases**: Quarto-rendered replication reports, academic blogs (`distill` \u002F `quarto blog`), graduate R courses, rigorous projects needing `marginaleffects` + `mediation` + `grf` post-estimation, anything R-flavoured outside of pure Bayesian work.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-24: 📊 Full Empirical Analysis Skill (Stata) shipped — traditional Stata ecosystem, 8-step .do loop (skills\u002F00.2)\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **📊 [Full Empirical Analysis Skill — Stata](skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002F)**: Same-day Stata sibling of StatsPAI \u002F 00.1, vendored at [`skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002F`](skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002F) — **slot #0.2, for Stata users**.\n  - **🧱 Traditional Stata ecosystem, de-facto standard command chain**: every step calls community-standard commands `reghdfe` \u002F `ivreg2` \u002F `ivreghdfe` \u002F `csdid` \u002F `did_imputation` \u002F `eventstudyinteract` \u002F `sdid` \u002F `did_multiplegt_dyn` \u002F `bacondecomp` \u002F `honestdid` \u002F `rdrobust` \u002F `rddensity` \u002F `synth` \u002F `synth_runner` \u002F `psmatch2` \u002F `teffects` \u002F `ebalance` \u002F `ppmlhdfe` \u002F `boottest` \u002F `ritest` \u002F `rwolf` \u002F `psacalc` \u002F `coefplot` \u002F `esttab` \u002F `outreg2` \u002F `asdoc` \u002F `binscatter` — **referee-level Stata replication packs, one `ssc install` block installs 30+ packages**.\n  - **🔁 8-step .do loop (same structure as 00.1, Stata-native rewrite)**: (1) Import & cleaning (`use`\u002F`import excel`\u002F`import sas`\u002F`destring`\u002F`misstable`\u002F`mdesc`\u002F`duplicates report`\u002F`merge m:1 ... assert(match using)`\u002F`xtset`\u002F`xtdescribe`\u002F`mi impute chained`) → (2) Variable construction (`winsor2 by(industry year)`\u002F`egen std`\u002F`xtile`\u002F`xtset` + `L.\u002FF.\u002FD.\u002FS.`\u002FCPI deflation\u002F`first_treat`+`rel_time`+`gvar`) → (3) Descriptives (`tabstat`\u002F`balancetable`\u002F`asdoc sum`\u002F`pwcorr, sig star(.05)`\u002F`heatplot`\u002F`twoway kdensity`\u002F`xtdescribe`) → (4) Diagnostics (12 classes: `swilk`\u002F`sktest`\u002F`estat hettest`\u002F`estat imtest, white`\u002F`xtserial`\u002F`xttest3`\u002F`xtcsd, pesaran`\u002F`estat vif`\u002F`dfuller`\u002F`kpss`\u002F`xtunitroot ips\u002Fllc`\u002F`hausman fe re`\u002F`estat ovtest`\u002F`linktest`) → (5) Estimation (12 classes: `reghdfe`+`areg`+`xtreg, fe\u002Fre`\u002F`ivreg2`+`ivreghdfe`+`ivregress liml\u002Fgmm`\u002F`csdid`+`eventstudyinteract`+`did_imputation`+`sdid`+`did_multiplegt_dyn`\u002F`rdrobust`+`rdmc`+`rddensity`\u002F`synth`+`synth_runner`\u002F`psmatch2`+`teffects psmatch\u002Fipwra\u002Faipw`+`ebalance`+`cem`\u002F`heckman`+`heckprob`\u002F`qreg`+`sqreg`\u002F`ppmlhdfe`\u002F`sem`+`gsem`) → (6) Robustness (`eststo`+`esttab` M1–M6, multi-cluster, `boottest`, `ritest`, `rwolf`, `bacondecomp`, `honestdid`, `psacalc delta`) → (7) Further analysis (factor-var interactions+`margins`+`marginsplot`\u002F`suest` cross-eq Wald\u002FDDD\u002Foutcome ladder coefplot\u002F`medsem`+`khb`+`sem` `estat teffects`\u002Fdose-response via `xtile` or `bspline`\u002FStata-Python bridge to `econml` for CATE\u002Fspillover) → (8) Publication output (`esttab`+`outreg2`+`asdoc` to `.tex`\u002F`.rtf`\u002F`.docx`\u002F`.xlsx`; `coefplot`+`marginsplot`+`binscatter`+`rdplot`+`graph combine` to `.pdf`).\n  - **📚 Progressive disclosure**: `SKILL.md` 801-line spine (full `ssc install` list + complete `.do` skeleton + library cheat-sheet); 8 [`references\u002FNN-*.md`](skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002Freferences\u002F) totalling 3500+ lines, loaded on demand.\n  - **🆚 Triple positioning** (now extended to 4 with 00.3): StatsPAI = Python DSL one-shot; 00.1 = explicit Python; 00.2 = **explicit Stata** — **the only choice when a referee or co-author insists on Stata replication**.\n  - **Use cases**: referee-level Stata replication packs, graduate Stata courses, AER\u002FQJE\u002FJPE\u002FReStud-style standard `.do` pipelines, rigorous research needing the full modern DID toolkit (`bacondecomp` + `honestdid` + `psacalc`).\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-24: 📘 Full Empirical Analysis Skill shipped — traditional Python econometric stack, explicit 8-step loop (skills\u002F00.1)\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **📘 [Full Empirical Analysis Skill](skills\u002F00.1-Full-empirical-analysis-skill_Python\u002F)**: Same-day sibling to StatsPAI, vendored at [`skills\u002F00.1-Full-empirical-analysis-skill_Python\u002F`](skills\u002F00.1-Full-empirical-analysis-skill_Python\u002F) — **slot #0.1, the explicit \u002F auditable counterpart**.\n  - **🧱 Traditional Python econometrics stack, no DSL wrapper**: every step directly calls `pandas` \u002F `numpy` \u002F `scipy` \u002F `statsmodels` \u002F `linearmodels` \u002F `pyfixest` \u002F `rdrobust` \u002F `econml` \u002F `causalml` \u002F `matplotlib` \u002F `seaborn` — every line of agent-written code is inspectable and swappable.\n  - **🔁 8-step closed loop (finer granularity than StatsPAI's 6 steps)**: (1) Data cleaning (MCAR\u002FMAR\u002FMNAR handling, IQR\u002Fz\u002FMahalanobis outliers, `validate=` on every merge, panel-structure checks) → (2) Variable construction (log\u002FIHS\u002FBox–Cox, 1\u002F99 winsorization, z\u002FMinMax\u002FRobust scaling, interactions\u002Flags\u002Fdiffs, CPI deflation, staggered-DID timing vars) → (3) Descriptive statistics (stratified Table 1 with SMDs+t-tests, starred correlation heatmap, 4-panel distribution figure, DID motivation plot, panel-coverage heatmap) → (4) Diagnostic tests (12 classes: normality \u002F heteroskedasticity \u002F autocorrelation \u002F multicollinearity \u002F stationarity \u002F cointegration \u002F endogeneity \u002F weak-IV \u002F overid \u002F panel Hausman \u002F RESET \u002F Cook's D) → (5) Baseline modeling (12 classes of estimators: OLS \u002F panel FE-RE-FD \u002F GLM \u002F IV-2SLS-LIML-GMM \u002F DID×5-2×2\u002FTWFE\u002Fevent-study\u002FCS\u002FSA\u002FBJS\u002FSDiD \u002F RD-Sharp\u002FFuzzy\u002FKink\u002Fmulti-cutoff \u002F SC \u002F PSM-IPW-EB \u002F DML \u002F Causal Forest \u002F Heckman \u002F Quantile) → (6) Robustness battery (M1–M6 progressive specs, cluster-level sensitivity, wild cluster bootstrap, placebo timing+permutation, specification curve, Oster δ\\*, LOO, Rosenbaum) → (7) Further analysis (heterogeneity × 4, outcome-ladder mechanism, Baron–Kenny + Imai mediation, moderated mediation, dose-response, spillover) → (8) Publication tables & figures (`stargazer` \u002F `pyfixest.etable` \u002F coefplot \u002F event-study \u002F binscatter \u002F forest \u002F RD plot \u002F CATE heatmap \u002F love plot, full LaTeX\u002FWord\u002FExcel export).\n  - **📚 Progressive-disclosure architecture**: `SKILL.md` holds only the one canonical call per step (610 lines of spine); variants are offloaded to 8 [`references\u002FNN-*.md`](skills\u002F00.1-Full-empirical-analysis-skill_Python\u002Freferences\u002F) deep manuals (3000+ lines total), **loaded by agents only when needed**.\n  - **🆚 Relationship to StatsPAI**: StatsPAI = **agent-native one-shot DSL** (one `sp.causal(...)` runs everything); this skill = **explicit traditional stack** (every line swappable, every diagnostic by hand). They coexist and complement — reach for StatsPAI when you trust the DSL; reach for this skill when teaching, auditing, or requiring full control.\n  - **Use cases**: replicating applied-economics papers, referee-level line-by-line audit, graduate teaching, any project that insists on hanging every diagnostic and robustness check into the explicit pipeline.\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-24: 🔥 StatsPAI Skill officially shipped — end-to-end automated empirical analysis (skills\u002F00)\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **🔥🔥 [StatsPAI Skill](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F)**: Our **agent-native, one-stop empirical-analysis Skill** is now officially vendored in this repo at [`skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F`](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F) — **slot #0, the repository's flagship**.\n  - **🚀 End-to-end automation for the entire empirical pipeline**: data cleaning (pandas pre-step) → EDA & descriptives (`sp.sumstats` \u002F `sp.balance_table`) → pre-flight diagnostics (`sp.diagnose` \u002F `sp.balance_panel` \u002F overlap \u002F missingness) → research-question DSL (`sp.causal_question(...).identify()`) → LLM-assisted DAG discovery (`sp.llm_dag_propose` \u002F `validate` \u002F `constrained`) → one-call estimation (`sp.causal(...)`) → robustness (`sp.spec_curve` \u002F `sp.honest_did` \u002F `sp.evalue`). **6-step closed loop, no tool switching — the agent runs the whole thing from a single instruction.**\n  - **900+ functions, one `import statspai as sp`**: more than doubled from the 390+ version on 2026-04-12. Covers OLS, IV, panel, DID (Callaway-Sant'Anna \u002F Sun-Abraham \u002F Bacon \u002F HonestDID \u002F continuous DID), RDD (Sharp \u002F Fuzzy \u002F multi-cutoff \u002F Kink), PSM, SCM, SDID, DML, Causal Forest, Meta-Learners, TMLE, AIPW, neural causal models (TARNet \u002F CFRNet \u002F DragonNet), **text causal (`sp.causal_text`)**, Heckman, structural estimation (BLP).\n  - **Agent-native self-describing API**: `sp.list_functions()` \u002F `sp.describe_function()` \u002F `sp.function_schema()` — agents discover and understand functions without doc lookup. Every estimator returns a unified `CausalResult` with `.summary()` \u002F `.plot()` \u002F `.to_latex()` \u002F `.to_word()` \u002F `.to_excel()` \u002F `.cite()` and a structured `.diagnostics` dict — **purpose-built for LLM-driven workflows**.\n  - **Estimand-first decisions**: `sp.causal_question` makes the \"DID vs RD vs IV?\" choice **explicit and defensible** — no more guesswork.\n  - **Submitted to JOSS, MIT-licensed.** [→ PyPI](https:\u002F\u002Fpypi.org\u002Fproject\u002FStatsPAI\u002F) | [→ GitHub](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI) | [→ Local Skill](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F)\n- **🔁 Weekly upstream sync**: new GitHub Action auto-pulls the latest `SKILL.md` \u002F `README.md` from the StatsPAI main repo into [`skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F`](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F) every week — **users always get the latest version**.\n- Corrected several `sp.*` signatures in Skill code examples; Step 0–6 code blocks are now explicitly flagged as *illustrative* (so agents don't copy them verbatim).\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-13: 🇨🇳 Original Chinese De-AIGC Skill Launched (skills\u002F48)\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **🇨🇳🔥 [chinese-de-aigc](skills\u002F48-copaper-ai-chinese-de-aigc\u002F)**: **CoPaper.AI team's original Chinese academic de-AIGC skill**. Currently the only humanizer on GitHub dedicated to Chinese academic empirical papers and targeting China's CNKI AMLC \u002F Wanfang \u002F VIP \u002F Turnitin Chinese detectors.\n  - **17-pattern library of Chinese AI tells** (4-character clichés \u002F hollow connectives \u002F explicit transitions \u002F absolutist claims \u002F total-part-total symmetry \u002F sentence-length uniformity)\n  - **5-step closed-loop workflow**: Locate → Diagnose → Differential Rewrite → 5-Dim Self-Score → Second-Pass Review\n  - **Per-section strategy**: Abstract \u002F Introduction \u002F Literature Review \u002F Methods \u002F Results \u002F Discussion \u002F Conclusion each has different rewrite intensity\n  - **5-dimension scoring rubric**: Concreteness \u002F Rhythm \u002F Caution \u002F Implicit Cohesion \u002F Researcher Voice (weighted max 50)\n  - **12 before\u002Fafter case comparisons** covering 7 main chapters of empirical papers\n  - Architecture inspired by English humanizers (humanizer_academic \u002F skill-deslop \u002F stop-slop \u002F avoid-ai-writing), **but fully re-designed for Chinese language context**\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-12: Added StatsPAI Agent-Native Econometrics Package + Anti-AIGC Detection Skills\u003C\u002Fb>\u003C\u002Fsummary>\n\n- **🔥 [StatsPAI](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI)**: Our own **agent-native causal inference & econometrics Python package**. 390+ functions, one `import`, self-describing API (`list_functions()` \u002F `describe_function()` \u002F `function_schema()`). Covers OLS, IV, DID (Callaway-Sant'Anna \u002F Sun-Abraham \u002F Bacon \u002F HonestDID \u002F continuous DID), RDD, PSM, SCM, DML, Causal Forest, Meta-Learners, TMLE, neural causal models (TARNet\u002FCFRNet\u002FDragonNet), and more. Published in JOSS, MIT license. [→ PyPI](https:\u002F\u002Fpypi.org\u002Fproject\u002FStatsPAI\u002F) | [→ GitHub](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI)\n- **📝 Anti-AIGC Detection Skills** (4 new, [→ dedicated section](#-anti-aigc-detection--de-ai-academic-writing-highlighted)):\n  - [humanizer_academic](https:\u002F\u002Fgithub.com\u002Fmatsuikentaro1\u002Fhumanizer_academic) — Academic paper specialist, 23 AI writing pattern detectors (`skills\u002F44`)\n  - [skill-deslop](https:\u002F\u002Fgithub.com\u002Fstephenturner\u002Fskill-deslop) — Scientific writing de-AI, respects discipline conventions (`skills\u002F45`)\n  - [stop-slop](https:\u002F\u002Fgithub.com\u002Fhardikpandya\u002Fstop-slop) — 3-layer detection + 5-dimension scoring (`skills\u002F46`)\n  - [avoid-ai-writing](https:\u002F\u002Fgithub.com\u002Fconorbronsdon\u002Favoid-ai-writing) — Structured audit + rewrite + second-pass audit (`skills\u002F47`)\n- **🛡️ [revision-guard](https:\u002F\u002Fgithub.com\u002FShiyanW\u002Fai-revision-guard)**: Prevents AI over-refinement, limits revision rounds + 7-point homogenization checklist (community PR contribution)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>2026-04-11: Expanded from 43 collections to 119 repos, covering 23,000+ Skills\u003C\u002Fb>\u003C\u002Fsummary>\n\n- Added 76 GitHub repositories across 8 social science disciplines (economics, political science, sociology, psychology, education, public health, management, finance)\n- Added skill suites for finance, law, marketing, product management, education, public health\n- Added 13 academic data MCP servers (OpenAlex, Semantic Scholar, FRED, World Bank, etc.)\n- Added 11 multi-agent collaboration systems (Agent Laboratory, AI-Scientist-v2, etc.)\n- Added bilingual Chinese\u002FEnglish README\n\n\u003C\u002Fdetails>\n\n---\n\n## Table of Contents\n\n- [Start Here](#start-here)\n- [🆕 Changelog](#-changelog)\n- [What Can This List Do for You?](#what-can-this-list-do-for-you)\n- [Quick Lookup by Research Stage](#quick-lookup-by-research-stage)\n- **Skills by Category**\n  - [01 - Topic Selection & Research Design](docs\u002F01-选题与研究设计.md)\n  - [02 - Literature Search & Review](docs\u002F02-文献检索与综述.md)\n  - [03 - Paper Reading & Analysis](docs\u002F03-论文阅读与拆解.md)\n  - [04 - Data Collection & Cleaning](docs\u002F04-数据获取与清洗.md)\n  - [05 - Statistical Analysis & Causal Inference](docs\u002F05-统计分析与因果推断.md)\n  - [06 - Paper Writing](docs\u002F06-论文写作.md)\n  - [07 - Paper Revision & Polishing](docs\u002F07-论文修改与润色.md)\n  - [08 - Citation Management & Typesetting](docs\u002F08-引用管理与排版.md)\n  - [09 - Replication & Reproducible Research](docs\u002F09-论文复现与可复现研究.md)\n  - [10 - Peer Review Response & Defense](docs\u002F10-审稿回复与学术答辩.md)\n- [Comprehensive Skill Suites](#comprehensive-skill-suites)\n  - 🚨 [Anti-AIGC Detection & De-AI Academic Writing (Highlighted)](#-anti-aigc-detection--de-ai-academic-writing-highlighted)\n- [Multi-Agent Collaboration Systems](#multi-agent-collaboration-systems)\n- [Skill Aggregation Platforms & Discovery Tools](#skill-aggregation-platforms--discovery-tools)\n- [Learning Resources](#learning-resources)\n- [🛡️ Security Scan](#-security-scan)\n- [Contributing](#contributing)\n\n---\n\n## What Can This List Do for You?\n\nIf you do empirical research, you've probably experienced these scenarios:\n\n- You ask AI to run a DID, and it gives you the baseline regression and stops. You say \"parallel trends?\" — it adds one. \"Placebo test?\" — another one. \"Event study plot?\" — yet another. **Every time, it's like squeezing toothpaste.**\n- You finally finish a draft, but citations are a mess, with a few hallucinated references mixed in.\n- You want to replicate an identification strategy from a top journal, but the gap between understanding it and implementing it feels like a mountain.\n\n**The problem isn't that AI can't do it — it doesn't know what a complete workflow should include.**\n\nA Skill solves this: it's a **methodological playbook for AI**. With a Skill, AI knows \"running DID means first testing parallel trends, then baseline regression, then 4 robustness checks, then heterogeneity analysis, then mechanism analysis, with specific output formats at each step.\" You just say \"run a DID analysis\" and it follows the complete workflow.\n\nThis list helps you find the best Skills for every stage of the empirical research workflow.\n\n---\n\n## Quick Lookup by Research Stage\n\n> Not sure which Skill to use? Start from your current research stage:\n\n```\nTopic Ideation → Lit Search → Deep Reading → Research Design → Data Collection\n      │              │             │              │                │\n      ▼              ▼             ▼              ▼                ▼\n     01             02            03             01               04\n\nData Cleaning → Statistical Analysis → First Draft → Revision → Typesetting\n      │              │                    │            │            │\n      ▼              ▼                    ▼            ▼            ▼\n     04             05                   06           07           08\n\nReplication → Submission → Peer Review Response → Defense\n      │           │              │                   │\n      ▼           ▼              ▼                   ▼\n     09          10             10                  10\n```\n\n### One-Stop Solutions\n\nIf you don't want to pick Skills one by one, these solutions cover the full workflow:\n\n| Solution | Coverage | Highlights | Link |\n|----------|----------|------------|------|\n| **CoPaper.AI** | Data Analysis → Paper Writing | 20 built-in methodology Skills, multi-agent architecture, complete publication-quality empirical paper in 20 minutes | [copaper.ai](https:\u002F\u002Fcopaper.ai) |\n| **StatsPAI Skill** 🔥🔥 | **End-to-end automated empirical analysis** | **900+ functions, one `import statspai as sp`**. A single agent instruction runs the full 6-step loop: EDA → pre-flight → research-question DSL → DAG discovery → estimation → robustness. Agent-native self-describing API, covers OLS\u002FIV\u002FDID (incl. Callaway-Sant'Anna, Sun-Abraham, HonestDID, continuous DID)\u002FRDD\u002FPSM\u002FSCM\u002FDML\u002FCausal Forest\u002Fneural causal\u002Ftext causal, publication-ready output (Word\u002FExcel\u002FLaTeX) | [Local Skill](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F) · [GitHub](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FStatsPAI) |\n| **Full Empirical Analysis Skill** 📘 | **Traditional Python stack, explicit 8-step loop** (StatsPAI's philosophical counterpart) | No DSL wrapper — the agent drives `pandas` \u002F `statsmodels` \u002F `linearmodels` \u002F `pyfixest` \u002F `rdrobust` \u002F `econml` \u002F `causalml` \u002F `matplotlib` directly. Covers **data cleaning → variable construction → descriptive statistics → diagnostic tests → modeling → robustness → mechanism\u002Fheterogeneity\u002Fmediation\u002Fmoderation → publication-ready tables & figures**, one deep reference per step. Ideal for teaching, referee-level audit, and strict replication work that needs line-by-line control | [Local Skill](skills\u002F00.1-Full-empirical-analysis-skill_Python\u002F) |\n| **Full Empirical Analysis Skill — Stata** 📊 | **Traditional Stata `.do` 8-step loop** (the Stata sibling of 00.1) | `reghdfe` + `ivreg2` + `csdid` + `did_imputation` + `eventstudyinteract` + `sdid` + `rdrobust` + `synth` + `psmatch2` + `teffects` + `ebalance` + `boottest` + `ritest` + `rwolf` + `bacondecomp` + `honestdid` + `coefplot` + `esttab` + `outreg2` + `asdoc` + `binscatter`. Same 8 steps, from `use` \u002F `import` all the way to `.tex` \u002F `.rtf` tables + `.pdf` figures. The first choice for referee-level Stata replication packs | [Local Skill](skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002F) |\n| **Full Empirical Analysis Skill — R** 📗 | **tidyverse + fixest 8-step loop + Quarto rendering** (R sibling of 00.1 \u002F 00.2) | `dplyr` + `haven` + `fixest` + `did` + `synthdid` + `bacondecomp` + `HonestDiD` + `rdrobust` + `tidysynth` + `gsynth` + `MatchIt` + `WeightIt` + `cobalt` + `ebal` + `grf` + `DoubleML` + `mediation` + `lavaan` + `marginaleffects` + `modelsummary` + `kableExtra` + `gt` + `flextable` + `ggplot2` + `iplot` + `binsreg` + `cowplot`. All 8 steps fit in a single `.qmd`; `quarto render` produces a unified PDF\u002FHTML\u002FWord reproducibility report. | [Local Skill](skills\u002F00.3-Full-empirical-analysis-skill_R\u002F) |\n| **Claude Scholar** | Ideation → Submission | 25+ Skills covering the full research lifecycle, Zotero MCP integration | [GitHub](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar) |\n| **K-Dense Scientific Skills** | Cross-disciplinary Science | 140+ Skills, 28+ scientific databases, 55+ Python packages | [GitHub](https:\u002F\u002Fgithub.com\u002FK-Dense-AI\u002Fclaude-scientific-skills) |\n| **AI-Research-SKILLs** | AI\u002FML Research | 22 categories, 87 skills, full research cycle | [GitHub](https:\u002F\u002Fgithub.com\u002FOrchestra-Research\u002FAI-Research-SKILLs) |\n| **OpenClaw Medical Skills** | Biomedical\u002FPublic Health | **869 Skills**, epidemiology, clinical research, drug safety, biostatistics | [GitHub](https:\u002F\u002Fgithub.com\u002FFreedomIntelligence\u002FOpenClaw-Medical-Skills) |\n| **Agent Laboratory** | Fully Autonomous Research | Lit review → Experiments → Report, 84% reduction in research costs | [GitHub](https:\u002F\u002Fgithub.com\u002FSamuelSchmidgall\u002FAgentLaboratory) |\n\n---\n\n## Comprehensive Skill Suites\n\nThese repositories contain multiple Skills and typically cover several research stages:\n\n### Academic Research\n\n| Suite | Stars | # Skills | Key Features | Social Science Fit |\n|-------|-------|----------|-------------|-------------------|\n| [K-Dense-AI\u002Fclaude-scientific-skills](https:\u002F\u002Fgithub.com\u002FK-Dense-AI\u002Fclaude-scientific-skills) | 8,799 | 140+ | 28+ scientific databases (OpenAlex, PubMed), scientific-writing + literature-review + statistical-analysis | ⭐⭐⭐⭐ |\n| [Orchestra-Research\u002FAI-Research-SKILLs](https:\u002F\u002Fgithub.com\u002FOrchestra-Research\u002FAI-Research-SKILLs) | 3,637 | 87 | 22 categories, ML paper writing, LaTeX templates, citation verification | ⭐⭐⭐ |\n| [Imbad0202\u002Facademic-research-skills](https:\u002F\u002Fgithub.com\u002FImbad0202\u002Facademic-research-skills) | ~1,790 | Multiple | Full paper pipeline (research → write → review → revise → finalize), style calibration, hallucination detection | ⭐⭐⭐⭐ |\n| [Galaxy-Dawn\u002Fclaude-scholar](https:\u002F\u002Fgithub.com\u002FGalaxy-Dawn\u002Fclaude-scholar) | - | 25+ | Full research lifecycle: ideation → review → experiments → writing → peer review response, Zotero MCP | ⭐⭐⭐⭐⭐ |\n| [luwill\u002Fresearch-skills](https:\u002F\u002Fgithub.com\u002Fluwill\u002Fresearch-skills) | 209 | 3 | Research proposal generation, medical review writing, paper-to-slides, bilingual | ⭐⭐⭐⭐⭐ |\n| [lishix520\u002Facademic-paper-skills](https:\u002F\u002Fgithub.com\u002Flishix520\u002Facademic-paper-skills) | 22 | 2 | Strategist (7-dimension reviewer simulation) + Composer (systematic writing) | ⭐⭐⭐⭐ |\n| [Data-Wise\u002Fclaude-plugins](https:\u002F\u002Fgithub.com\u002FData-Wise\u002Fclaude-plugins) | - | 17 | Statistical research: arXiv search, DOI lookup, BibTeX management, methodology writing, referee response | ⭐⭐⭐⭐⭐ |\n\n### Economics \u002F Causal Inference\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| **[CoPaper.AI](https:\u002F\u002Fcopaper.ai)** | **20 methodology Skills** (OLS, DID, staggered DID, IV, RDD, PSM, SCM, DML, causal forest, etc.), multi-agent architecture (Supervisor + 4 sub-agents), smart routing, automatic output | Full empirical economics workflow |\n| **[StatsPAI Skill](skills\u002F00-Full-empirical-analysis-skill_StatsPAI\u002F)** 🔥🔥 | **End-to-end automated empirical analysis.** Agent-native econometrics Python package: **900+ functions**, one `import statspai as sp` runs the full loop: EDA → research-question DSL → LLM-assisted DAG discovery → estimation → robustness. Self-describing API (`list_functions()` \u002F `describe_function()` \u002F `function_schema()`), unified `CausalResult` objects. Covers OLS, IV, panel data, DID (Callaway-Sant'Anna \u002F Sun-Abraham \u002F Bacon \u002F HonestDID \u002F continuous DID), RDD (Sharp\u002FFuzzy\u002Fmulti-cutoff\u002FKink), PSM, SCM, SDID, DML, Causal Forest, Meta-Learners, TMLE, AIPW, neural causal models (TARNet\u002FCFRNet\u002FDragonNet), **text causal (`sp.causal_text`)**, Heckman, structural estimation (BLP). **Submitted to JOSS, MIT license** | Whole-pipeline automation: one agent call goes from cleaned data to robust estimates |\n| **[Full Empirical Analysis Skill](skills\u002F00.1-Full-empirical-analysis-skill_Python\u002F)** 📘 | **Traditional Python econometrics stack, explicit 8-step closed loop** (philosophical counterpart to StatsPAI: DSL one-shot vs. explicit line-by-line). No wrapper — drives `pandas` + `numpy` + `scipy` + `statsmodels` + `linearmodels` + `pyfixest` + `rdrobust` + `econml` + `causalml` + `matplotlib` + `seaborn` directly. Fine-grained 8 steps: (1) data cleaning (MCAR\u002FMAR\u002FMNAR, IQR\u002Fz\u002FMahalanobis, `validate=` safe merges, panel-structure checks) → (2) variable construction (log\u002FIHS\u002FBox–Cox, 1\u002F99 winsorization, z\u002FMinMax\u002FRobust, interactions\u002Flags\u002Fdiffs, CPI deflation, staggered-DID timing) → (3) descriptives (stratified Table 1 with SMD+t-tests, starred correlation heatmap, 4-panel distributions, DID motivation plot, panel-coverage heatmap) → (4) diagnostics (12 classes: normality \u002F heteroskedasticity \u002F autocorrelation \u002F collinearity \u002F stationarity \u002F cointegration \u002F endogeneity \u002F weak-IV \u002F overid \u002F Hausman \u002F RESET \u002F Cook's D) → (5) modeling (OLS \u002F panel FE-RE-FD \u002F GLM \u002F IV-2SLS-LIML-GMM \u002F 5 DID variants \u002F 4 RD variants \u002F SC \u002F PSM-IPW-EB \u002F DML \u002F CF \u002F Heckman \u002F QR — 12 classes) → (6) robustness (M1–M6 progressive specs, cluster sensitivity, wild bootstrap, placebo, spec curve, Oster δ\\*, LOO, Rosenbaum) → (7) further analysis (heterogeneity × 4 \u002F outcome-ladder mechanism \u002F Baron–Kenny + Imai mediation \u002F moderated mediation \u002F dose-response \u002F spillover) → (8) publication tables & figures (`stargazer` \u002F `etable` \u002F coefplot \u002F event-study \u002F binscatter \u002F forest \u002F RD plot \u002F CATE heatmap \u002F love plot, plus LaTeX\u002FWord\u002FExcel export). **610-line SKILL.md spine + 8 deep reference manuals (3000+ lines), progressively loaded** | Teaching, referee-level audit, graduate replication training, rigorous empirical projects requiring line-by-line control and full diagnostic coverage |\n| **[Full Empirical Analysis Skill — Stata](skills\u002F00.2-Full-empirical-analysis-skill_Stata\u002F)** 📊 | **Traditional Stata `.do` 8-step closed loop** (Stata sibling of 00.1, same structure, same cadence). One `ssc install` block installs 30+ packages. End-to-end community-standard chain: `reghdfe` \u002F `ivreg2` \u002F `ivreghdfe` \u002F `csdid` \u002F `did_imputation` \u002F `eventstudyinteract` \u002F `sdid` \u002F `did_multiplegt_dyn` \u002F `bacondecomp` \u002F `honestdid` \u002F `rdrobust` \u002F `rddensity` \u002F `synth` \u002F `synth_runner` \u002F `psmatch2` \u002F `teffects` \u002F `ebalance` \u002F `ppmlhdfe` \u002F `boottest` \u002F `ritest` \u002F `rwolf` \u002F `psacalc` \u002F `coefplot` \u002F `esttab` \u002F `outreg2` \u002F `asdoc` \u002F `binscatter`. 8 steps: (1) `use`+`import`+`destring`+`misstable`+`merge assert`+`xtset` → (2) `winsor2`+`xtile`+`L.\u002FF.\u002FD.\u002FS.`+CPI+staggered timing → (3) `tabstat`+`balancetable`+`asdoc`+`pwcorr sig star`+`heatplot` → (4) 12 estat-style diagnostics → (5) 12 estimator classes (`reghdfe` + 5 DID + 4 RD + `synth` + `teffects` + `ebalance` + `heckman` + `qreg` + `ppmlhdfe` + `sem\u002Fgsem`) → (6) `eststo`+`esttab` M1–M6 + `boottest` + `ritest` + `rwolf` + `bacondecomp` + `honestdid` + `psacalc delta` → (7) factor-var + `margins` + `marginsplot` + `suest` + DDD + `medsem` + `khb` + SEM + Stata-Python bridge to `econml` for CATE → (8) `esttab`+`outreg2`+`asdoc` to `.tex\u002F.rtf\u002F.docx\u002F.xlsx`; `coefplot`+`marginsplot`+`binscatter`+`rdplot`+`graph combine` to `.pdf`. **801-line SKILL.md + 8 deep references (3500+ lines) + complete `.do` skeleton** | Referee \u002F co-author insists on Stata replication; graduate Stata courses; AER\u002FQJE\u002FJPE\u002FReStud-style standard `.do` pipelines |\n| **[Full Empirical Analysis Skill — R](skills\u002F00.3-Full-empirical-analysis-skill_R\u002F)** 📗 | **Modern tidyverse + fixest + Quarto stack, explicit 8-step loop** (R sibling of 00.1 \u002F 00.2; the fourth and final piece of the family). One `install.packages(...)` block installs 50+ packages. End-to-end modern R standards: `dplyr` \u002F `tidyr` \u002F `haven` \u002F `janitor` \u002F `naniar` \u002F `mice` \u002F `validate` \u002F `assertr` for data; `fixest::feols\u002Ffeglm\u002Ffepois` for HD FE + multi-way clustering + IV in one line; `did::att_gt` \u002F `fixest::sunab` \u002F `didimputation::did_imputation` \u002F `synthdid` \u002F `DIDmultiplegtDYN` \u002F `bacondecomp` \u002F `HonestDiD` for modern DID; `rdrobust` \u002F `rddensity` \u002F `rdmulti` \u002F `rdlocrand` for RD; `Synth` \u002F `gsynth` \u002F `tidysynth` \u002F `synthdid` for SC; `MatchIt` \u002F `WeightIt` \u002F `cobalt` \u002F `ebal` for matching; `grf::causal_forest` \u002F `DoubleML` for ML causal; `mediation::mediate` + `medsens` \u002F `lavaan::sem` for mediation; `marginaleffects` for post-estimation; `modelsummary` \u002F `kableExtra` \u002F `gt` \u002F `flextable` for tables; `ggplot2` + `iplot` + `binsreg` + `cowplot` + `patchwork` for figures. 8-step R pipeline + **Quarto template** (one `.qmd` holding narrative + code + tables + figures, `quarto render` for PDF\u002FHTML\u002FWord in one go). **893-line SKILL.md + 8 deep references (3700+ lines)**, progressively loaded | Quarto reproducibility reports, academic blogs (distill \u002F quarto blog), graduate R courses, projects needing `marginaleffects` + Imai sensitivity mediation + `grf` CATE post-estimation |\n| **[AER-Skills](skills\u002F50-brycewang-aer-skills\u002F)** 📕🔥 | **Top-5 economics submission skill stack** (AER \u002F AER:Insights \u002F AEJ family), complementary to StatsPAI \u002F 00.x \"run the analysis\" — specialised in \"write the paper + submit + R&R\". **Nine skills, full pipeline**: `aer-topic-selection` (AER vs Insights vs AEJ routing) → `aer-identification` (identification audit: modern DiD \u002F weak IV \u002F boundary RDD pitfalls) → `aer-robustness` (referee-anticipating robustness matrix) → `aer-introduction` (Keith Head five-paragraph intro) → `aer-tables-figures` (AER booktabs typesetting) → `aer-replication` (AEA Data and Code Availability Policy package, openICPSR-ready) → `aer-submission` (preflight: 100-word abstract, disclosure, cover letter) → `aer-rebuttal` (R&R letters written against the *revised* manuscript) → `aer-workflow` (orchestrator). **Identification-first** — if your design is fragile, no prose will save it. Covers AER 100-word abstract \u002F AER:Insights 7000-word limit \u002F 45% desk-rejection \u002F AEA mandatory replication — top-5-specific constraints that generic scientific-writing skills do not cover. **[`scripts\u002Fsync-aer-skills.sh`](scripts\u002Fsync-aer-skills.sh) + weekly GH Actions loop syncs from upstream [brycewang-stanford\u002FAER-skills](https:\u002F\u002Fgithub.com\u002Fbrycewang-stanford\u002FAER-skills)**. **License: MIT** | Full AER \u002F AER:Insights \u002F AEJ submission flow: topic routing → identification audit → writing → typesetting → replication package → submission → R&R rebuttal |\n| [claesbackman\u002FAI-research-feedback](https:\u002F\u002Fgithub.com\u002Fclaesbackman\u002FAI-research-feedback) | 2-agent economics paper pre-review: causal overclaiming detection, identification strategy assessment; supports AER\u002FQJE\u002FJPE\u002FEconometrica\u002FREStud; 6-agent grant review | Pre-submission self-review, grant applications |\n| [fuhaoda\u002Fstats-paper-writing-agent-skills](https:\u002F\u002Fgithub.com\u002Ffuhaoda\u002Fstats-paper-writing-agent-skills) | LaTeX statistical paper writing, front-end draft generation | Statistics & econometrics papers |\n| [dylantmoore\u002Fstata-skill](https:\u002F\u002Fgithub.com\u002Fdylantmoore\u002Fstata-skill) | Full Stata coverage: syntax, data management, econometrics, causal inference, graphics, Mata, 20+ community packages | Stata users |\n| [SepineTam\u002Fstata-mcp](https:\u002F\u002Fgithub.com\u002FSepineTam\u002Fstata-mcp) | LLM operates Stata regression directly via MCP, \"evolve from regression monkey to causal thinker\" | Stata econometrics |\n\n### 🚨 Anti-AIGC Detection & De-AI Academic Writing (Highlighted)\n\n> **This is one of the most critical pain points in academic writing in 2026**. Papers failing AIGC detection can be rejected outright, and detectors like Turnitin, GPTZero, and China's CNKI are getting stricter. The 4 skills below are the **most authoritative and complete** solutions on GitHub — all MIT open-source, and all locally archived in this repo (`skills\u002F44-47`).\n\n| Suite | Key Features | Use Case | Local Path |\n|-------|-------------|----------|-----------|\n| **CoPaper.AI \u002F chinese-de-aigc** 🇨🇳🔥 | **Original Chinese academic de-AIGC skill** by CoPaper.AI team. Targets China's CNKI AMLC \u002F Wanfang \u002F VIP \u002F Turnitin Chinese detectors. 17-pattern library of Chinese-specific AI tells (4-char clichés, hollow connectives, explicit transitions, absolutist claims, sentence-length uniformity), 5-step closed loop workflow (locate→diagnose→rewrite→self-score→review), per-section strategy, 5-dim scoring rubric. **Currently the only GitHub skill dedicated to Chinese academic de-AIGC** | Chinese journal submissions, theses, grant proposals | [`skills\u002F48`](skills\u002F48-copaper-ai-chinese-de-aigc\u002F) |\n| **[matsuikentaro1\u002Fhumanizer_academic](https:\u002F\u002Fgithub.com\u002Fmatsuikentaro1\u002Fhumanizer_academic)** 🔥 | **Academic-specific**. 23 AI writing patterns (6 content + 6 language + 3 style + 3 filler + 5 word choice), examples from EMPA-REG OUTCOME cardiovascular trials, preserves legitimate academic transitions, based on Wikipedia \"Signs of AI writing\" | Medical, life sciences, natural science papers | [`skills\u002F44`](skills\u002F44-matsuikentaro1-humanizer_academic\u002F) |\n| **[stephenturner\u002Fskill-deslop](https:\u002F\u002Fgithub.com\u002Fstephenturner\u002Fskill-deslop)** | **Scientific writing de-AI**. Smartly distinguishes legitimate discipline conventions (passive voice in methods) from AI tells; 5-dimension scoring (directness\u002Frhythm\u002Ftrust\u002Fauthenticity\u002Fdensity); 4 reference files (examples\u002Fphrases\u002Fstructures\u002Ftropes) | Scientific papers, technical blogs | [`skills\u002F45`](skills\u002F45-stephenturner-skill-deslop\u002F) |\n| **[hardikpandya\u002Fstop-slop](https:\u002F\u002Fgithub.com\u002Fhardikpandya\u002Fstop-slop)** | **3-layer detection + 5-dim scoring**. Banned phrases (throat-clearing openers, emphasis crutches, corporate jargon), structural clichés (binary contrasts, dramatic fragmentation, false agency), sentence-level rules (no em dash, no Wh- starters). Below 35\u002F50 → revise | General prose, blogs, reports | [`skills\u002F46`](skills\u002F46-hardikpandya-stop-slop\u002F) |\n| **[conorbronsdon\u002Favoid-ai-writing](https:\u002F\u002Fgithub.com\u002Fconorbronsdon\u002Favoid-ai-writing)** | **Structured audit + rewrite + second-pass audit**. Four-section output: identified issues (with quotes) → rewrite → change summary → second audit. Compatible with Claude Code, OpenClaw, Hermes, and other agents | Workflows needing auditable, traceable revision | [`skills\u002F47`](skills\u002F47-conorbronsdon-avoid-ai-writing\u002F) |\n| [ShiyanW\u002Fai-revision-guard](https:\u002F\u002Fgithub.com\u002FShiyanW\u002Fai-revision-guard) | **Prevents over-refinement** (different angle). Limits revision rounds (≤2 per section), 7-point homogenization checklist, cross-model verification. Protects author's voice from AI erosion | Multi-round polishing scenarios | (community PR) |\n\n> **Recommended combos**:\n> - 🇨🇳 **Chinese academic papers** (CNKI\u002FWanfang\u002FVIP) → **chinese-de-aigc** (original) + **revision-guard**\n> - 🇬🇧 English academic papers → **humanizer_academic** + **revision-guard** (prevent over-refinement)\n> - Bilingual papers → **chinese-de-aigc** + **humanizer_academic** combined\n> - Need auditable workflow → **avoid-ai-writing** (structured reports)\n> - General writing → **stop-slop** (5-dim scoring for quantified improvement)\n\n### Finance & Investment Research\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| [anthropics\u002Ffinancial-services-plugins](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Ffinancial-services-plugins) | Anthropic official: investment banking, equity research, private equity, wealth management | Financial services |\n| [OctagonAI\u002Fskills](https:\u002F\u002Fgithub.com\u002FOctagonAI\u002Fskills) | Octagon agentic financial research Claude Skills | Institutional financial research |\n| [tradermonty\u002Fclaude-trading-skills](https:\u002F\u002Fgithub.com\u002Ftradermonty\u002Fclaude-trading-skills) | Stock investing & trading: market analysis, technical charts, economic calendar, strategy development | Quantitative trading research |\n| [himself65\u002Ffinance-skills](https:\u002F\u002Fgithub.com\u002Fhimself65\u002Ffinance-skills) | Agent Skills open standard, earnings analysis, consensus estimates, analyst sentiment | Financial analysis |\n| [quant-sentiment-ai\u002Fclaude-equity-research](https:\u002F\u002Fgithub.com\u002Fquant-sentiment-ai\u002Fclaude-equity-research) | Institutional equity research: fundamental analysis, technical indicators, risk assessment | Equity research |\n\n### Education & Public Health\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| [GarethManning\u002Fclaude-education-skills](https:\u002F\u002Fgithub.com\u002FGarethManning\u002Fclaude-education-skills) | Evidence-based education Claude Skills, designed for teachers and agent orchestration | Education research |\n| [FreedomIntelligence\u002FOpenClaw-Medical-Skills](https:\u002F\u002Fgithub.com\u002FFreedomIntelligence\u002FOpenClaw-Medical-Skills) | **869** medical AI Skills: epidemiology, public health surveillance, clinical research, drug safety, biostatistics | Public health, medical research |\n\n### Governance, Compliance & Law\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| [Sushegaad\u002FClaude-Skills-Governance-Risk-and-Compliance](https:\u002F\u002Fgithub.com\u002FSushegaad\u002FClaude-Skills-Governance-Risk-and-Compliance) | GRC Skills: ISO 27001, SOC 2, GDPR, HIPAA compliance guidance (94% vs 72% baseline) | Compliance research, policy analysis |\n| [zubair-trabzada\u002Fai-legal-claude](https:\u002F\u002Fgithub.com\u002Fzubair-trabzada\u002Fai-legal-claude) | Legal assistant: contract review, risk analysis, NDA generation, compliance audit, 14 Skills + 5 agents | Law & economics, regulatory research |\n| [evolsb\u002Fclaude-legal-skill](https:\u002F\u002Fgithub.com\u002Fevolsb\u002Fclaude-legal-skill) | AI contract review: CUAD risk detection, market benchmarks, attorney-grade red-lining | Law & economics research |\n\n### Marketing & Consumer Behavior\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| [coreyhaines31\u002Fmarketingskills](https:\u002F\u002Fgithub.com\u002Fcoreyhaines31\u002Fmarketingskills) | CRO, copywriting, SEO, analytics, and growth engineering | Marketing research |\n| [zubair-trabzada\u002Fai-marketing-claude](https:\u002F\u002Fgithub.com\u002Fzubair-trabzada\u002Fai-marketing-claude) | 15 Skills + parallel sub-agents: website audit, copy, email sequences, competitive intelligence | Consumer behavior analysis |\n| [ericosiu\u002Fai-marketing-skills](https:\u002F\u002Fgithub.com\u002Fericosiu\u002Fai-marketing-skills) | Growth experiments, sales pipeline, content operations, SEO, financial automation | Marketing strategy research |\n\n### Product Management & Organizational Behavior\n\n| Suite | Key Features | Use Case |\n|-------|-------------|----------|\n| [phuryn\u002Fpm-skills](https:\u002F\u002Fgithub.com\u002Fphuryn\u002Fpm-skills) | 100+ agent Skills: discovery → strategy → execution → launch → growth, 65 PM Skills + 36 chained workflows | Product management, organizational research |\n| [mastepanoski\u002Fclaude-skills](https:\u002F\u002Fgithub.com\u002Fmastepanoski\u002Fclaude-skills) | UX\u002FUI evaluation (Nielsen heuristics, WCAG), AI governance (NIST AI RMF, ISO 42001) | UX research |\n\n### General Agent Capabilities\n\n| Suite | Stars | Key Features |\n|-------|-------|-------------|\n| [lyndonkl\u002Fclaude](https:\u002F\u002Fgithub.com\u002Flyndonkl\u002Fclaude) | - | 85 skills + 6 orchestration agents, incl. causal inference, Bayesian reasoning, experimental design, multi-criteria analysis |\n| [alirezarezvani\u002Fclaude-skills](https:\u002F\u002Fgithub.com\u002Falirezarezvani\u002Fclaude-skills) | ~5,200 | 220+ skills + 298 CLI scripts, incl. financial analysis and data processing |\n| [rohitg00\u002Fawesome-claude-code-toolkit](https:\u002F\u002Fgithub.com\u002Frohitg00\u002Fawesome-claude-code-toolkit) | - | 135 agents incl. data scientist agent (EDA, DID, RDD), 35 skills, 42 commands |\n| [jeremylongshore\u002Fclaude-code-plugins-plus-skills](https:\u002F\u002Fgithub.com\u002Fjeremylongshore\u002Fclaude-code-plugins-plus-skills) | - | 340 plugins + **1,367 agent skills**, CCPI package manager |\n| [affaan-m\u002Feverything-claude-code](https:\u002F\u002Fgithub.com\u002Faffaan-m\u002Feverything-claude-code) | - | Skills, intuition, memory, security, research-first development framework |\n| [posit-dev\u002Fskills](https:\u002F\u002Fgithub.com\u002Fposit-dev\u002Fskills) | - | Posit official: modern-r-tidyverse, predictive-modeling, quarto-authoring, shiny-bslib |\n\n---\n\n## Multi-Agent Collaboration Systems\n\nA single Skill solves a point problem; multi-agent systems solve **end-to-end workflows**. These systems let multiple AI roles divide work, cross-review, and produce output quality far beyond what a single agent can achieve:\n\n### Paper Revision & Writing\n\n| System | Architecture | Key Features |\n|--------|-------------|-------------|\n| **copy-edit-master** | 3 sub-agents: structure-editor + line-editor + quality-reviewer | Auto document type detection, Strunk & White \u002F McCloskey rules encoded, git checkpoints per phase, review loop (max 2 iterations) |\n| **introduction-writer** | 4 sub-agents: strategist → drafter → reviewer → reviser | Keith Head formula for drafting introductions, reviewer independent from drafter for quality loop |\n| **CoPaper.AI PaperAgent** | Supervisor + 4 sub-agents (preparation \u002F modeling \u002F visualization \u002F writing) | Skills routed by target_agent, each sub-agent sees only relevant methodology guidance, reduced context noise |\n\n> **Why multi-agent beats single agent?** When the same agent writes and reviews, it tends to approve its own work. Role separation means the reviewer is independent from the drafter — forming a genuine quality loop. Same logic as academic peer review.\n\n### Data Analysis & Research\n\n| System | Source | Key Features |\n|--------|--------|-------------|\n| [ruc-datalab\u002FDeepAnalyze](https:\u002F\u002Fgithub.com\u002Fruc-datalab\u002FDeepAnalyze) | Renmin Univ. | Autonomous data analysis agent, raw data → professional report, CSV\u002FExcel\u002FJSON\u002FDB support, open-source DeepAnalyze-8B |\n| [business-science\u002Fai-data-science-team](https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fai-data-science-team) | Business Science | Multi-agent data science team: EDA Agent + SQL Agent + MLflow Agent, LangChain integration |\n| [HungHsunHan\u002Fclaude-code-data-science-team](https:\u002F\u002Fgithub.com\u002FHungHsunHan\u002Fclaude-code-data-science-team) | Community | Claude Code multi-agent data science team, auto cleaning → modeling → executable Notebook |\n| [HKUDS\u002FAI-Researcher](https:\u002F\u002Fgithub.com\u002FHKUDS\u002FAI-Researcher) | HKU (NeurIPS 2025 Spotlight) | Fully autonomous research pipeline: lit review → hypothesis → algorithm → paper |\n| [wanshuiyin\u002FAuto-claude-code-research-in-sleep (ARIS)](https:\u002F\u002Fgithub.com\u002Fwanshuiyin\u002FAuto-claude-code-research-in-sleep) | Community | Overnight autonomous research, cross-model review loops (Claude + external LLM as critic) |\n| [SamuelSchmidgall\u002FAgentLaboratory](https:\u002F\u002Fgithub.com\u002FSamuelSchmidgall\u002FAgentLaboratory) | Academic (ICLR) | End-to-end autonomous research: lit review → experiments → report, arXiv\u002FHuggingFace\u002FLaTeX integration, 84% cost reduction |\n| [SakanaAI\u002FAI-Scientist-v2](https:\u002F\u002Fgithub.com\u002FSakanaAI\u002FAI-Scientist-v2) | Sakana AI | Fully automated scientific discovery: hypothesis → experiments → paper, first AI-generated paper accepted via peer review |\n| [assafelovic\u002Fgpt-researcher](https:\u002F\u002Fgithub.com\u002Fassafelovic\u002Fgpt-researcher) | Community | Autonomous deep research agent, supports any LLM provider |\n| [LitLLM\u002FLitLLM](https:\u002F\u002Fgithub.com\u002FLitLLM\u002FLitLLM) | Academic | AI literature review assistant: keyword extraction + multi-strategy retrieval + re-ranking, RAG-based |\n| [pedrohcgs\u002Fclaude-code-my-workflow](https:\u002F\u002Fgithub.com\u002Fpedrohcgs\u002Fclaude-code-my-workflow) | Emory Univ. | Academic LaTeX\u002FBeamer + R template, multi-agent review + quality gates, adopted by 15+ research groups |\n| [hugosantanna\u002Fclo-author](https:\u002F\u002Fgithub.com\u002Fhugosantanna\u002Fclo-author) | Community | Extends Sant'Anna's workflow from lecture production to full social science empirical research publication |\n\n### Academic Data MCP Servers\n\n| System | Key Features |\n|--------|-------------|\n| [xingyulu23\u002FAcademix](https:\u002F\u002Fgithub.com\u002Fxingyulu23\u002FAcademix) | Unified academic research interface aggregating OpenAlex + DBLP + Semantic Scholar + arXiv + CrossRef |\n| [Eclipse-Cj\u002Fpaper-distill-mcp](https:\u002F\u002Fgithub.com\u002FEclipse-Cj\u002Fpaper-distill-mcp) | 11-source parallel search, 4-dimension weighted ranking (relevance\u002Frecency\u002Fimpact\u002Fnovelty) |\n| [oksure\u002Fopenalex-research-mcp](https:\u002F\u002Fgithub.com\u002Foksure\u002Fopenalex-research-mcp) | OpenAlex API: search 240M+ academic works, citation analysis, trend tracking, collaboration networks |\n| [zongmin-yu\u002Fsemantic-scholar-fastmcp-mcp-server](https:\u002F\u002Fgithub.com\u002Fzongmin-yu\u002Fsemantic-scholar-fastmcp-mcp-server) | Full Semantic Scholar API access: papers, authors, citation networks |\n| [openags\u002Fpaper-search-mcp](https:\u002F\u002Fgithub.com\u002Fopenags\u002Fpaper-search-mcp) | Search 20+ sources: arXiv, PubMed, bioRxiv, Google Scholar, SSRN, Unpaywall, etc. |\n| [aringadre76\u002Fmcp-for-research](https:\u002F\u002Fgithub.com\u002Faringadre76\u002Fmcp-for-research) | Integrates PubMed + Google Scholar + ArXiv + JSTOR, published on NPM |\n| [blazickjp\u002Farxiv-mcp-server](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server) | arXiv paper search and analysis MCP |\n| [lzinga\u002Fus-gov-open-data-mcp](https:\u002F\u002Fgithub.com\u002Flzinga\u002Fus-gov-open-data-mcp) | 40+ US government APIs (FRED\u002FCensus\u002FCDC\u002FFDA\u002FFEC, etc.), 250+ tools |\n| [stefanoamorelli\u002Ffred-mcp-server](https:\u002F\u002Fgithub.com\u002Fstefanoamorelli\u002Ffred-mcp-server) | Direct access to FRED's 800K+ economic time series |\n| [llnOrmll\u002Fworld-bank-data-mcp](https:\u002F\u002Fgithub.com\u002Fllnormll\u002Fworld-bank-data-mcp) | World Bank Data360, 1000+ socioeconomic indicators, 200+ countries |\n| [54yyyu\u002Fzotero-mcp](https:\u002F\u002Fgithub.com\u002F54yyyu\u002Fzotero-mcp) | Connect Zotero library with AI assistants: paper review, summaries, citation analysis, PDF annotation |\n| [datagouv\u002Fdatagouv-mcp](https:\u002F\u002Fgithub.com\u002Fdatagouv\u002Fdatagouv-mcp) | French national open data platform MCP |\n\n---\n\n## Skill Aggregation Platforms & Discovery Tools\n\nDon't know where to find Skills? These platforms are your starting point:\n\n| Platform | Scale | Features |\n|----------|-------|----------|\n| [VoltAgent\u002Fawesome-agent-skills](https:\u002F\u002Fgithub.com\u002FVoltAgent\u002Fawesome-agent-skills) | 1,000+ skills | 13,700 stars, curated by official team and community |\n| [sickn33\u002Fantigravity-awesome-skills](https:\u002F\u002Fgithub.com\u002Fsickn33\u002Fantigravity-awesome-skills) | 1,340+ skills | 28,000 stars, one-click install `npx antigravity-awesome-skills` |\n| [VoltAgent\u002Fawesome-openclaw-skills](https:\u002F\u002Fgithub.com\u002FVoltAgent\u002Fawesome-openclaw-skills) | **5,400+ skills** | Curated from OpenClaw registry (ClawHub 13,729 Skills) |\n| [jeremylongshore\u002Fclaude-code-plugins-plus-skills](https:\u002F\u002Fgithub.com\u002Fjeremylongshore\u002Fclaude-code-plugins-plus-skills) | 1,367 skills | 340 plugins + CCPI package manager |\n| [skills.sh](https:\u002F\u002Fskills.sh\u002F) | Online market | Searchable Skill marketplace |\n| [ClawHub (clawhub.com)](https:\u002F\u002Fclawhub.com) | **13,729 skills** | Open-source AI skill marketplace, one-line install |\n| [Agent Skills Standard](https:\u002F\u002Fagentskills.io\u002F) | Spec docs | Universal Agent Skills specification |\n| [Anthropic Official Skills](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills) | Official | PDF\u002FDOCX\u002FXLSX\u002FPPTX document processing |\n| [Anthropic Official Plugin Market](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official) | Official | Anthropic-managed high-quality Claude Code plugin catalog |\n| [Anthropic Knowledge Work Plugins](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fknowledge-work-plugins) | Official | 11 plugins incl. Data Plugin (SQL queries, data exploration) |\n| [Anthropic Financial Services Plugins](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Ffinancial-services-plugins) | Official | Financial services plugins: IB, equity research, PE, wealth mgmt |\n\n---\n\n## Learning Resources\n\n### Official Documentation\n\n- [Claude Code Skills Complete Guide](https:\u002F\u002Fresources.anthropic.com\u002Fhubfs\u002FThe-Complete-Guide-to-Building-Skill-for-Claude.pdf) — Anthropic's official 32-page guide\n- [Agent Skills Standard Specification](https:\u002F\u002Fagentskills.io\u002F)\n- [Claude Code Official Docs](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fdocs\u002Fagents-and-tools\u002Fagent-skills)\n\n### Academic Talks & Courses\n\n- [AI Agents for Economics Research](https:\u002F\u002Fcepr.org\u002F) — Aniket Panjwani, CEPR\u002FVoxDev, 2026.03\n- [Claude Code & Cowork for Academic Research — A Practical Guide](https:\u002F\u002Fcornwl.github.io\u002Ffiles\u002Fclaude-academic-guide.html) — Practical guide for economists and social scientists, 2026.02\n- [Building Claude Code Workflow for Economics Scholars](https:\u002F\u002Fzhiyuanryanchen.github.io\u002Fclaude-code-workflow.html) — Building Claude Code workflows for economics researchers\n\n### Causal Inference Textbooks\n\n- [Causal Inference for the Brave and True](https:\u002F\u002Fgithub.com\u002Fxieliaing\u002FCausalInferenceIntro) — Chinese translation, Python code\n- [Statistical Tools for Causal Inference](https:\u002F\u002Fchabefer.github.io\u002FSTCI\u002F) — Open-source textbook\n- [Causal Inference and Machine Learning Book](https:\u002F\u002Fwww.causalmlbook.com\u002F)\n\n### Survey Papers & Awesome Lists\n\n- [A Survey of Data Agents](https:\u002F\u002Fgithub.com\u002FHKUSTDial\u002Fawesome-data-agents) — Data agent survey (HKUST)\n- [From AI for Science to Agentic Science](https:\u002F\u002Fgithub.com\u002FAgenticScience\u002FAwesome-Agent-Scientists) — arXiv:2508.14111\n- [From Automation to Autonomy](https:\u002F\u002Fgithub.com\u002FHKUST-KnowComp\u002FAwesome-LLM-Scientific-Discovery) — LLM scientific discovery survey (EMNLP 2025)\n- [Awesome Agents for Science](https:\u002F\u002Fgithub.com\u002FOSU-NLP-Group\u002Fawesome-agents4science) — Papers on LLMs and agents in scientific R&D\n- [Awesome AI for Science](https:\u002F\u002Fgithub.com\u002Fai-boost\u002Fawesome-ai-for-science) — AI tools, papers, datasets for accelerating scientific discovery\n- [Awesome AI for Economists](https:\u002F\u002Fgithub.com\u002Fhanlulong\u002Fawesome-ai-for-economists) — AI tools, libraries, an","Auto-Empirical-Research-Skills 是一个精选的AI Agent技能库，包含超过23,000个技能，覆盖8大社会科学学科的实证研究。该项目能够自动运行从数据清洗到结果呈现的完整实证研究流程，包括识别与估计、稳健性检验以及生成图表和论文草稿等步骤，极大减少了人工干预的需求。主要使用Stata语言编写，适合需要进行高效且可复现的社会科学实证研究场景，如经济学、政治学、心理学等领域。由斯坦福大学REAP中心与CoPaper.AI联合维护。",2,"2026-06-11 04:08:48","high_star"]