Claude Data Analysis Assistant

A modern, intelligent data analysis platform built with Claude Code's sub-agents, slash-commands, skills, and hooks. Transform your data analysis workflow with AI-powered assistance and specialized analysis tools.

简单的一句话： 2个命令， /do-all 常规数据分析； /do-more 互联网数据分析。而分析数据是放在 /data_storage 。就这么简单，用起来吧！

注意：下载项目下来，分析数据是放在 /data_storage [删去原来的demo数据] ，你需要先删除 complete_analysis 和 do_more_analysis 这两个文件夹。我这里放着是给你参考最终的分析结果，作为例子。

🚀 Quick Start

1. Set Up Your Data

Place your dataset in the data_storage/ directory:

cp your_data.csv ./data_storage/

2. Start Analysis

Use intuitive slash commands to analyze your data:

# Complete interactive workflow with human feedback checkpoints
/do-all

# ⭐ NEW: Automatic multi-skill analysis
/do-more

# Basic exploratory analysis
/analyze user_behavior_sample.csv exploratory

# Create visualizations
/visualize user_behavior_sample.csv all

# Generate analysis code
/generate python data-cleaning

# Create comprehensive report
/report user_behavior_sample.csv complete markdown

🎯 Key Features

⭐ /do-more vs /do-all: Which Should You Use?

`/do-more`: Automatic Multi-Skill Analysis

Best for: Quick, automated analysis without configuration

/do-more  # No parameters needed!

What it does:

✅ Automatically scans data_storage/ directory
✅ Identifies data types (e-commerce, user behavior, etc.)
✅ Intelligently matches 7+ relevant skills
✅ Executes skills in optimal order
✅ Generates comprehensive HTML report
✅ No human intervention required
✅ Fast execution (2-5 minutes)

Output: do_more_analysis/integrated_results/Comprehensive_Analysis_Report.html

`/do-all`: Complete Interactive Analysis Workflow

Best for: Thorough analysis with human oversight and feedback

/do-all

What it does:

✅ Reads data from data_storage/ (no parameters needed!)
✅ 6-stage workflow with quality checks
✅ 3 Human feedback checkpoints at critical stages
✅ Interactive hypothesis generation
✅ Custom code generation
✅ Comprehensive documentation
✅ Multiple output formats (HTML, PDF, Markdown, DOCX)

Workflow Stages:

Data Quality Assessment → ⚠️ [human checkpoint #1] - Confirm data quality
Exploratory Analysis - Statistical summaries, patterns, trends
Hypothesis Generation → ⚠️ [human checkpoint #2] - Review research directions
Visualization → ⚠️ [human checkpoint #3] - Approve visualization strategy
Code Generation - Reproducible analysis pipeline
Report Generation - Comprehensive final report

Output Directory:

complete_analysis/
├── data_quality_report/          # Stage 1 output
├── exploratory_analysis/         # Stage 2 output
├── hypothesis_reports/           # Stage 3 output
├── visualizations/               # Stage 4 output
├── generated_code/               # Stage 5 output
├── final_report/                 # Stage 6 output
└── workf...

claude-data-analysis-ultra-main