👥 Team Research & Collaboration

Share Claude Code setups across your lab, use Git for reproducible research, and standardise workflows so your whole team benefits from AI assistance.

Git + Claude Code: the basics

Here's a minimal Git workflow for research projects:

# 1. Initialise Git in your project (do this once)
git init
git add .
git commit -m "Initial project setup"

# 2. Before a Claude Code session
git status                    # See what's changed

# 3. After Claude makes changes you want to keep
git add -A
git commit -m "Claude: cleaned survey dataset"

# 4. If Claude made a mistake and you want to undo
git checkout -- analysis.do   # Undo changes to one file
git reset --hard HEAD         # Undo ALL changes since last commit

Sharing your Claude Code setup

Because everything is just files, sharing your Claude Code setup is automatic when you share your project via Git. Here's what gets shared:

Team CLAUDE.md best practice

# Add to your shared CLAUDE.md for team projects:

## Team Conventions
- All team members use Stata 17 or later
- Code review required before merging to main branch
- All analysis scripts must be runnable from scratch (no local paths)
- Use English for all code comments and variable names
- Ask before modifying files in data/raw/

## Contact
- PI: [name] — ask before major structural changes
- RA: [name] — handles data cleaning do files
- For questions about the analysis: see analysis/README.md

Reproducibility standards

> I want to make this project fully reproducible so anyone can
  replicate our results from scratch. Please:
  1. Check all file paths — flag any that are hardcoded to my computer
  2. Document all software versions needed (Stata 17, packages, etc.)
  3. Create a "README_replication.md" with step-by-step instructions
  4. Check that the analysis can run in the correct order with main.do
  5. Identify any data that can't be shared (PII) and suggest what to
     include instead (e.g., simulated data with same structure)

Working with RAs and co-authors

Briefing a new RA

> A new research assistant is joining our project tomorrow.
  They know Stata but haven't seen this project before.
  Create an onboarding document (docs/RA_onboarding.md) that:
  - Explains what the project is about (2 paragraphs)
  - Describes the folder structure and what's in each folder
  - Lists the 5 most important do files and what they do
  - Explains our coding conventions from CLAUDE.md
  - Gives them 3 starter tasks to get familiar with the data

Code review with Claude

> My RA wrote this data cleaning script (analysis/01_clean.do).
  Review it for:
  1. Correctness: will it produce the right output?
  2. Efficiency: are there unnecessary steps?
  3. Conventions: does it follow our CLAUDE.md style rules?
  4. Robustness: will it break if the data structure changes?
  Give specific, constructive feedback she can act on.