菜单
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.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| 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"] |
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.
.claude/rules/domain-profile.md for field conventionsquality_reports/ (if analysis implements a pre-specified design)CLAUDE.md for language preference (R/Stata/Python/Julia)code/build/ and code/analysis/ for project patternsDelegate 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:
fixest for panel data, modelsummary for tables, ggplot2 for figures.tex tables for LaTeX, .pdf/.png figures, .rds for intermediate objectsAfter 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
If Debugger finds Critical or Major issues:
three-strikes.md)Show the user:
output/NN_issue/tables/, figures in output/NN_issue/figures/# ============================================================
# [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 ----
Fetch structured data from external APIs for research workflows. Use when Codex needs to query a documented API, save raw responses reproducibly, transform them into project-ready tables, or document authentication and rate-limit constraints for later reruns.
Compile LaTeX papers or slide decks reproducibly. Use when Codex needs to build a paper or Beamer deck, summarize warnings, rerun bibliography steps, or verify that a manuscript target compiles cleanly before review or submission.
Summarize current session state and preserved project context. Use when Codex needs to report what files and plans capture the current state, identify what will survive a long session boundary, or point the user to the latest durable artifacts on disk.
Prepare data and code materials for deposit or archival handoff. Use when Codex needs to check repository contents against deposit requirements, organize metadata, verify provenance notes, or prepare deposit-facing documentation without changing the scientific content.
Deploy or publish project artifacts safely. Use when Codex needs to prepare a site, service, documentation artifact, or generated output for release, confirm prerequisites, and outline the reproducible deployment steps without skipping verification.
Implement or review econometrics workflows in Julia. Use when Codex needs to write Julia estimation code, check design-to-code alignment, structure reproducible Julia scripts, or adapt an empirical workflow into Julia while preserving the project's numbered analysis conventions.