Quick Start

Get running in 5 minutes

Research papers have a production function: identify, estimate, write, submit. This scaffold automates each stage with specialized agents that create, review, and gate-check every artifact before it advances. One Claude Code configuration handles the full pipeline — from literature review to journal submission.

Works for empirical economics out of the box. Adapts to labor, public, health, development, trade, IO, and other applied fields by customizing the domain profile and journal profiles.

What to Expect

When you run a command, Claude enters plan mode, dispatches specialized agents, scores output against quality gates, and presents results for your approval. This is not autocomplete — it is an orchestrated pipeline where every artifact is created by one agent and reviewed by another before advancing.

The pipeline at a glance:

  1. Discovery — literature search and data assessment, producing an annotated bibliography and ranked dataset list.
  2. Strategy — identification design, producing a strategy memo with assumptions, pseudo-code, and robustness plan.
  3. Execution — analysis code and paper manuscript, producing scripts, tables, figures, and LaTeX sections.
  4. Review — simulated peer review, producing scored referee reports and an editorial decision.
  5. Submission — replication packaging and quality gate enforcement (aggregate score >= 95, all components >= 80).

You can enter at any stage. Each phase produces structured output saved to quality_reports/.

Step 1: Fork & Clone

gh repo fork hugosantanna/clo-author --clone
cd clo-author

Step 2: Start Claude Code

claude

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

Step 3: Paste the Starter Prompt

Fill in the bolded placeholders with your project details:

I am starting a new empirical research project in [YOUR FIELD] on [YOUR TOPIC]. Read CLAUDE.md and help me set up the project structure. Start with /discover lit [YOUR TOPIC].

Claude reads the configuration, fills in your project details, and begins planning — it will present a plan for your approval before doing anything.

Step 4: Start Working

Once setup is done, describe what you need:

  • “Review my paper” — checks identification strategy, code, writing, and LaTeX in parallel
  • /review --methods paper/main.tex — validates your causal design assumptions, clustering, and inference
  • /analyze Data/cleaned/survey.csv — cleans data, runs your specification, produces tables and figures
  • /new-project health insurance — walks you through the full pipeline from literature review to submission
TipJust Want to Try 2 Commands?

Already have a paper? /review paper/main.tex runs writer-critic, strategist-critic, and the verifier. Want to draft a section? /write intro produces a structured first draft adapted to your paper type. Everything else is optional — explore when you’re ready.

Guide Navigation

Page What you will find
User Guide Common workflows: first review, first draft, full pipeline, R&R, talks
Agents What each of the 19 agents does, checks, and assumes
Rules & Invariants The 22 content invariants, severity gradient, and quality gates
Hooks & Automation Event-driven scripts: file protection, linting, context survival
Command Reference Every command, its flags, agents dispatched, and output artifacts
Architecture Dependency graph, scoring formula, lifecycle validation
Customization Adapting the scaffold to your field, journal, and preferences

That’s it. Everything else — agents, hooks, rules — runs automatically in the background.

If you prefer to configure things yourself instead of letting Claude handle it:

Customize CLAUDE.md — Open CLAUDE.md and replace all [BRACKETED PLACEHOLDERS]:

  1. Project name and institution
  2. Folder structure (adjust to your layout)
  3. Current project state (your papers and analyses)

Fill in the domain profile — Run /discover (interactive interview) or manually edit .claude/references/domain-profile.md:

  1. Your field and adjacent subfields
  2. Target journals ranked by tier
  3. Common data sources and identification strategies
  4. Field conventions and seminal references

Configure permissions — Review .claude/settings.json. The template includes permissions for git, LaTeX, R, and utility scripts.

Test it:

/tools compile paper/main.tex
/review --proofread paper/main.tex

Prerequisites

Tool Required For Install
Claude Code Everything npm install -g @anthropic-ai/claude-code
XeLaTeX Paper compilation TeX Live or MacTeX
R Analysis & figures r-project.org
gh CLI GitHub integration brew install gh (macOS)

Optional: Python, Julia (multi-language analysis), Quarto (web slides).

Origin

Maintained by Hugo Sant’Anna. MIT License — fork it, customize it, make it yours.