data-analysis

name	data-analysis
description	End-to-end data analysis dispatching the Coder agent for implementation and Debugger agent for code review. Supports R, Stata, Python, and Julia. Produces scripts, tables, figures, and results summary.
disable-model-invocation	true
argument-hint	[dataset path or description of analysis goal]
allowed-tools	["Read","Grep","Glob","Write","Edit","Bash","Task"]

Data Analysis Workflow

Run an end-to-end data analysis by dispatching the Coder (implementer) and Debugger (code critic).

Input: $ARGUMENTS — a dataset path (e.g., data/build/01_clean/panel.csv) or a description of the analysis goal.

---

Workflow

Step 1: Context Gathering

1. Read .claude/rules/domain-profile.md for field conventions
2. Read any strategy memo in quality_reports/ (if analysis implements a pre-specified design)
3. Check CLAUDE.md for language preference (R/Stata/Python/Julia)
4. Scan existing scripts in code/build/ and code/analysis/ for project patterns

Step 2: Launch Coder Agent

Delegate to the coder agent via Task tool:

Prompt: Implement analysis for "[goal]" using data at "[path]".
Follow 4 stages:
  Stage 0: Data cleaning (if raw data provided)
  Stage 1: Main specification (from strategy memo or user description)
  Stage 2: Robustness checks
  Stage 3: Publication-ready output (tables to output/NN_issue/tables/, figures to output/NN_issue/figures/)
Produce results_summary.md with all estimates, SEs, and key statistics.
Save scripts to code/build/NN_issue/ or code/analysis/NN_issue/ (numbered by GitHub issue).

The Coder follows these principles:

• Script structure: Header, setup, data loading, analysis, output, export
• Packages: fixest for panel data, modelsummary for tables, ggplot2 for figures
• Standard errors: Cluster at appropriate level (match treatment assignment)
• Output: .tex tables for LaTeX, .pdf/.png figures, .rds for intermediate objects
• No hardcoded paths. All paths relative to repository root.

Step 3: Launch Debugger Agent (Code Critic)

After Coder returns, delegate to the debugger agent:

Prompt: Review the script(s) at code/analysis/NN_issue/[script_name].R.
Run all 12 check categories:
  Strategic (1-3): code-strategy alignment, sanity checks, robustness sufficiency
  Code Quality (4-12): structure, console hygiene, reproducibility, functions,
    figure quality, RDS pattern, comments, error handling, polish
If strategy memo exists, cross-reference code against stated design.
Save report to quality_reports/[script]_code_review.md

Step 4: Fix Issues

If Debugger finds Critical or Major issues:

1. Re-dispatch Coder with specific fixes (max 3 rounds per three-strikes.md)
2. Re-run Debugger to verify fixes

Step 5: Present Results

Show the user:

1. Results summary — key estimates with SEs and interpretation
2. Scripts created — paths and descriptions
3. Output files — tables in output/NN_issue/tables/, figures in output/NN_issue/figures/
4. Code review score — from Debugger
5. Any TODO items — missing data, additional specifications needed

---

Script Structure Template

# ============================================================
# [Descriptive Title]
# Author: [from project context]
# Purpose: [What this script does]
# Inputs: [Data files]
# Outputs: [Figures, tables, RDS files]
# ============================================================

# 0. Setup ----
library(tidyverse)
library(fixest)
library(modelsummary)

set.seed(42)

dir.create("Tables", recursive = TRUE, showWarnings = FALSE)
dir.create("Figures", recursive = TRUE, showWarnings = FALSE)

# 1. Data Loading ----

# 2. Exploratory Analysis ----

# 3. Main Analysis ----

# 4. Tables and Figures ----

# 5. Export ----

---

Principles

• Reproduce, don't guess. If the user specifies a regression, run exactly that.
• Show your work. Print summary statistics before jumping to regressions.
• Strategy alignment. If a strategy memo exists, the code MUST implement it faithfully.
• Worker-critic pairing. Coder creates, Debugger critiques. Never skip the review.
• saveRDS everything. Every computed object gets saved for downstream use.
• Publication-ready output. Tables and figures should be directly includable in the paper.