felpix-research

Social Science Research

Work in Progress. This plugin is still under active development by a university student. For now, it is an experiment on the utility of Claude Code for social science research. If you have any feedback or suggestions, reach me on X at @felpix_.

A Claude Code plugin based on Pedro Sant'Anna's Claude Code workflow designed for producing social science research.

Quick Start

1. Install

Use Claude Code to install the plugin in your project directory.

/plugin marketplace add Felpix-Studios/social-science-research

/plugin install social-science-research@felpix-research

Why project scope? The plugin creates quality_reports/, templates/, references/, and manuscripts/ directories on every session start. Installing at user (global) scope would scaffold those folders in every project you open — including non-research repos. Project scope keeps everything contained to the repo you're working in.

2. Restart Claude Code

In your project directory, on first launch, the plugin will automatically:

• Create project directories: quality_reports/, references/papers/, manuscripts/, output/, scripts/
• Copy references/domain-profile.md into your project (your field's journals, datasets, and key researchers)
• Copy starter templates into templates/ (session log, requirements spec, quality report formats)
• Create CLAUDE.md with project identity placeholders (name, author, institution)

3. Run /research-setup

Configure your field, institution, journals, datasets, key researchers, and institutional colors. This writes your domain profile and customizes R figure themes to match your institution's visual identity.

Enjoy using the plugin!

Prerequisites

Tool	Required For	Install
Claude Code (with plugin support)	Everything	claude.ai/download
R (>= 4.0)	/data-analysis (R track), /review-r	r-project.org
Python (>= 3.9)	/data-analysis (Python track), compact hooks	python.org
LaTeX distribution	/write-paper with .tex output	tug.org/texlive

How It Works

Step	What You Do	Skill(s)	Output
0. Setup	Configure field, journals, datasets, R colors	/research-setup	references/domain-profile.md, CLAUDE.md
1. Idea	Interview → ideation → unified project spec	/new-project	quality_reports/project_spec_*.md
2a. Literature	Parallel journal + working paper + citation search	/lit-review then /validate-bib	quality_reports/lit_review_*.md
2b. Data	Find and assess datasets for the research question	/data-finder	quality_reports/data_exploration_*.md
3. Analysis	Run R or Python analysis, review code quality	/data-analysis then /review-r	output/tables/, output/figures/, scripts/
4. Write	Draft and review the manuscript	/write-paper then /review-paper	manuscripts/[name]-draft.tex
5. Verify	Check analysis-paper consistency, proofread	/quality-gate then /proofread	quality_reports/quality_gate_*.md

What's Included

7 agents, 12 skills, 8 rules, 4 hooks (click to expand)

Agents

Agent	What It Does
librarian	Search one literature angle and return BibTeX
explorer	Find datasets across one source category
explorer-critic	Stress-test dataset candidates on 5 dimensions
proofreader	Grammar, typos, layout, and consistency check
domain-reviewer	Top-journal referee review through 5 lenses
r-reviewer	R code quality and reproducibility review
verifier	Trace paper claims to output files

Skills

Skill	What It Does
/research-setup	Configure field, institution, journals, datasets, and colors
/new-project	Interactive interview to formalize a research idea
/lit-review	Literature search, synthesis, and gap identification
/validate-bib	Cross-reference citations against bibliography
/data-finder	Find and rank datasets for a research question
/data-analysis	End-to-end R or Python analysis with publication-ready output
/review-r	Launch R code reviewer on scripts
/write-paper	Draft manuscript from analysis outputs
/review-paper	Manuscript review: structure, econometrics, referee objections
/quality-gate	Verify every paper claim traces to an output file
/proofread	Launch proofreader on a file
/deep-audit	Repository-wide consistency audit

Rules