Social Science Research Skills for Claude Code

A curated collection of Claude Code skills that encode rigorous social science methodology directly into your AI coding workflow. These skills transform Claude from a general-purpose coding assistant into a methodologically aware research partner that enforces statistical best practices, prevents common analytical errors, and produces publication-quality outputs.

What Are Claude Code Skills?

Skills are markdown files that give Claude Code deep domain expertise. When you place a SKILL.md file in a .claude/skills/ directory, Claude automatically loads the relevant skill based on your task -- enforcing best practices, catching methodological errors, and generating code that follows disciplinary standards.

Unlike generic prompts, skills encode mandatory requirements with consequences. For example, the Difference-in-Differences skill won't let Claude generate a DID analysis without parallel trends testing -- because Bertrand, Duflo & Mullainathan (2004) showed that skipping this step leads to a 45% false positive rate. Every requirement cites the research that justifies it.

Skills Included

Foundation

Skill	Description
Core Methodology	Loaded on every analysis. Enforces reproducibility (seeds, version documentation), defensive input validation, RATIONALE comments on every statistical choice, and transparent data provenance. Built on lessons from the replication crisis -- Nosek et al. (2015) found only 39% of psychology studies replicated, often due to undocumented analytical choices.

Causal Inference

Skill	Description
Difference-in-Differences	The workhorse of policy evaluation. Enforces parallel trends testing, cluster-robust standard errors, event study specifications, and proper handling of staggered adoption (Callaway & Sant'Anna 2021). Includes pre-trends diagnostic templates.
RCT Design	Randomized controlled trial design from pre-registration through theory of change. Requires pre-analysis plans, adequate power, IRB consideration, and clear identification of treatment mechanisms.
RCT Randomization	Implementation of random assignment with seed documentation, permutation-based methods, stratification, and cluster randomization. Produces verifiable, reproducible allocation sequences.
RCT Power Calculations	Statistical power analysis with design effect adjustments for clustering, attrition inflation factors, and minimum detectable effect size computation. Prevents underpowered studies that waste resources.
RCT Data Analysis	Intent-to-treat analysis, LATE estimation, balance tables, attrition handling, and multiple testing corrections. Enforces cluster-robust standard errors and pre-specified outcome hierarchies.
RCT Ethics & Registration	IRB protocols, informed consent frameworks, trial registration (AEA RCT Registry, ClinicalTrials.gov), pre-analysis plan templates, and data security requirements.

Analysis & Communication