Skip to content

CLAUDE-template.md

Latest commit

 

History

History
68 lines (49 loc) · 3.34 KB

CLAUDE-template.md

File metadata and controls

68 lines (49 loc) · 3.34 KB

Academic Research Project Instructions for Claude

Working Directory Context

You are working in the ProjectExample/ folder, which is a Git repository. This folder contains:

  • Git-tracked folders: Code/, Figures/, Tables/, Paper/, Slides/
  • Symlinked folders: Data, Notes, Output (these link to ../ProjectExample-Share/ which is outside the Git repo)

You can access all folders normally - the symlinks are transparent. Files in symlinked folders are NOT tracked by Git but are synced via Dropbox.

Project Structure

This project follows a two-folder structure designed for academic research collaboration:

Git Repository (ProjectExample/)

  • Code/ - All analysis scripts organized by task (e.g., DataCleaning/)
  • Figures/ - Final presentable charts and visualizations (version-tracked)
  • Tables/ - Final presentable results and summary statistics (version-tracked)
  • Paper/ - LaTeX documents for academic papers
  • Slides/ - LaTeX presentations
  • Data, Notes, Output - Symlinks to corresponding folders in ProjectExample-Share/

Dropbox Folder (ProjectExample-Share/)

  • Data/ - Raw and processed datasets (read-only, not version-tracked)
  • Notes/ - Research notes and documentation
  • Output/ - Intermediate results organized by task matching Code/ structure

Subfolder Organization

Code/ Directory

Organize scripts by research tasks, not by individual runs. Examples:

  • Code/DataCleaning/ - Scripts for data preparation and cleaning
  • Code/Analysis/ - Main analysis scripts
  • Code/Robustness/ - Robustness checks and sensitivity analyses
  • Code/Visualization/ - Scripts generating figures and tables

Important: Edit existing scripts rather than creating new ones for variations of the same task.

Output/ Directory

Mirror the Code/ structure with corresponding output folders:

  • Output/DataCleaning/ - Cleaned datasets, processing logs
  • Output/Analysis/ - Regression results, statistical outputs
  • Output/Robustness/ - Alternative specification results
  • Output/Visualization/ - Draft figures and tables (not final versions)

Within each Output subfolder, optionally organize by script name:

  • Output/Analysis/main_regression.py/ - Outputs from main_regression.py
  • Output/Analysis/heterogeneity.py/ - Outputs from heterogeneity.py

Coding Style

  • Code is for academic research and NOT meant for production-ready. Therefore, write concise and efficient code without safety check (try...catch..., if...else) unless it's necessary or specifically requested
  • Due to the exploratory nature, DO write interactive code that can be evaluated line-by-line
  • Document only when necessary, but be concise
  • When producing outputs, save in Output/ following the subfolder convention. Do NOT save outputs in Figures/ or Tables/ unless explicitly requested
  • The project is version controlled with Git. Hence, when adding new analysis, Do NOT create a new script per task, but DO edit the existing files directly
  • Always execute at the project root rather than cd into subfolders

Python Environment

  • Uses uv for dependency management
  • Virtual environment located at ~/.venvs/ProjectExample
  • Run commands with uv run <command> (e.g., uv run python script.py)
  • Add dependencies with uv add package

Whenever calling Python-related programs, use uv unless it is infeasible.