TECH BLOG

例）
xxxxx /             # スキルのフォルダ = スキル名
├── SKILL.md        # スキルの中身（主にAIへの指示手順が書かれている）
└── scripts/        # スキル発動で参考にするスクリプトのフォルダ
    └── xxxxx.py    # 参考スクリプト（エージェントが参照）

---
name: csv-analyzer
description: Analyze CSV data files and generate a structured business
  report with summary statistics, charts, and actionable insights.
  Use this skill when users upload CSV or tabular data and ask for
  analysis or reporting.
---

# CSV Data Analyzer

Analyze uploaded CSV data and produce a business-ready report.

## Workflow

1. **Load data**: Read the CSV file using pandas.
   Show the shape (rows x columns) and first 5 rows.
2. **Basic statistics**: Compute summary stats
   (mean, median, min, max) for all numeric columns.
3. **Identify trends**: Find the top/bottom performers,
   detect outliers, and note any patterns.
4. **Generate chart**: Create a bar chart or line chart of
   the most important metric using matplotlib.
   Save to `output/chart.png`.
5. **Write report**: Output a structured Markdown report with
   sections: Overview, Key Findings, Chart, and Recommendations.

## Rules
- Always respond in the same language the user uses
- If data has Japanese column headers, keep them as-is
- Handle encoding issues (try utf-8, then shift_jis)
- If no CSV is provided, generate sample sales data to demonstrate

#!/usr/bin/env python3
"""CSV Analyzer - Generate business report from CSV data."""
import sys, os, json
import pandas as pd
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

def generate_sample_data():
    data = {
        'month': ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun'],
        'sales': [120, 150, 130, 170, 160, 190],
        'cost': [80, 90, 85, 100, 95, 110],
        'region': ['Tokyo', 'Tokyo', 'Osaka', 'Tokyo', 'Osaka', 'Osaka']
    }
    df = pd.DataFrame(data)
    df.to_csv('sample_sales.csv', index=False)
    return df, 'sample_sales.csv'

def analyze(filepath):
    for enc in ['utf-8', 'shift_jis', 'cp932']:
        try:
            df = pd.read_csv(filepath, encoding=enc)
            break
        except:
            continue
    else:
        return "Error: Could not read CSV"

    os.makedirs('output', exist_ok=True)
    stats = df.describe()
    numeric = df.select_dtypes(include='number')

    if not numeric.empty:
        col = numeric.columns[0]
        df.plot(kind='bar', y=col, x=df.columns[0], figsize=(10, 6))
        plt.title(f'{col} Analysis')
        plt.tight_layout()
        plt.savefig('output/chart.png', dpi=100)

    print(json.dumps({
        'shape': list(df.shape),
        'columns': list(df.columns),
        'stats': stats.to_dict(),
        'head': df.head().to_dict()
    }, ensure_ascii=False, default=str))

if __name__ == '__main__':
    if len(sys.argv) > 1:
        analyze(sys.argv[1])
    else:
        df, path = generate_sample_data()
        print(f"Generated sample: {path}")
        analyze(path)