AI-Assisted Academic Research Template (Make + R/Julia/Stata/MATLAB)

Work in progress. This is a summary of how I use AI coding assistants for computational research — running analysis pipelines with Make, writing R, Julia, Stata, and MATLAB scripts, and managing build dependencies. I keep updating these files as I learn new things.

A ready-to-fork starter kit for researchers using Claude Code, OpenAI Codex CLI, or Kimi Code CLI with Make + R + Julia + Stata + MATLAB build systems. You describe what you want; the assistant selects the appropriate risk tier, implements and tests routine work directly, and adds plans, independent review, or durable context only when the task warrants them.

All three tools are supported. Claude Code uses a CLAUDE.md hierarchy plus .claude/; Codex CLI uses an AGENTS.md hierarchy plus .codex/ and .agents/; Kimi Code CLI uses the same AGENTS.md hierarchy, scans .agents/skills/ natively, and mirrors permissions through .kimi-code/config.toml.example. The same workflow, quality gates, and skills work with each.

---

Quick Start: Claude Code (5 minutes)

1. Fork & Clone

# Fork this repo on GitHub (click "Fork" on the repo page), then:
git clone https://github.com/YOUR_USERNAME/repo-template.git my-project
cd my-project

Replace YOUR_USERNAME with your GitHub username.

2. Start Claude Code and Paste This Prompt

claude

Using VS Code? Open the Claude Code panel instead. Everything works the same.

Then paste the following, filling in your project details:

I am starting to work on [PROJECT NAME] in this repo. [Describe your project in 2–3 sentences — what you’re building, what data you use, what your pipeline stages are (e.g., data cleaning, estimation, figures).]

I want our collaboration to be structured, precise, and rigorous. Code should be reproducible, build-system driven, and written for clarity over cleverness.

I’ve set up the Claude Code academic workflow (forked from irudik/repo-template). The configuration files are already in this repo. Please read them, understand the workflow, and then update all configuration files to fit my project — fill in placeholders in CLAUDE.md, set up the code/ directory structure with sub-Makefiles for my pipeline stages, and propose any customizations specific to my use case.

After that, use the risk-based workflow: implement and test routine changes directly; use a brief in-conversation plan for substantive single-module work; and save a plan before high-risk, cross-cutting, numerical, or pre-merge work, then continue automatically unless I explicitly request plan approval.

Start adapting the workflow configuration for this project.

Optional for Claude Code + Codex plugin users: Add this to the prompt when