Agentic AI for Social Science Research

A Tutorial Series on Claude Code for Empirical Workflows

Author

Bogdan G. Popescu

Published

March 24, 2026

Preface

This tutorial series teaches empirical social science researchers how to use Claude Code — an agentic AI harness — for the full research workflow: from reading papers and designing identification strategies through analysis, verification, and journal submission.

The series is organized in four parts:

  1. Introduction — What agentic AI is, the expertise-verification paradox, and how to install Claude Code
  2. CLAUDE.md — Configuring Claude Code for research with a persistent project constitution
  3. Git — Version control as safety infrastructure for AI-assisted workflows
  4. Skills — A reusable skills library for the publication pipeline, from literature discovery through submission

Each tutorial is self-contained but they build on each other. The accompanying skills_library/ folder contains 13 working skills you can install into your own projects.

The tutorials synthesize ideas from Scott Cunningham’s MixtapeTools, Hugo Sant’Anna’s Clo-Author, Chris Blattman’s claudeblattman.com, Aniket Panjwani’s AI MBA, and Anthropic’s Claude Code documentation.