📊 CSV Tools v1.1.0

Overview

12个子命令覆盖日常CSV数据处理的完整工作流——从预览、筛选、排序到统计、类型检测、数据画像和抽样。所有操作基于Python标准库csv模块，无需安装任何第三方依赖。

Usage

stats — 列统计 (NEW)

python3 scripts/csv_tools.py stats data.csv

# 📊 Column Statistics — data.csv
#    [amount] — 数值列
#      计数: 1000 | 空: 0 | 总和: 485,230.00
#      平均: 485.23 | 中位: 320.00
#      最小: -8,500.00 | 最大: 25,000.00 | 标准差: 1,234.56
#    [name] — 文本列
#      非空: 980 | 空: 20 | 唯一值: 450
#      最常见: "张三" (15次)

columns — 列操作 (NEW)

# 重命名 + 选择列
python3 scripts/csv_tools.py columns data.csv \
  --rename "old_name=new_name" --select "id,name,amount"

# 添加计算列
python3 scripts/csv_tools.py columns data.csv \
  --add "total=price * quantity" --output result.csv

detect — 类型检测 (NEW)

python3 scripts/csv_tools.py detect --verbose data.csv

# 🔎 Data Type Detection — data.csv
#    [amount] → 浮点数(Float)
#    [date] → 日期(Date)
#    [is_active] → 布尔(Boolean)
#    [name] → 文本

profile — 数据画像 (NEW)

python3 scripts/csv_tools.py profile data.csv

# 📋 Data Profile — data.csv
#    Rows: 1000 | Columns: 8
#    [email]
#      非空: 850 / 1000 (████████████████████ 15% 缺失)
#      唯一值: 830 (98%)
#      最常见: "user@example.com"(3), ...

sample — 随机抽样 (NEW)

# 随机抽10条
python3 scripts/csv_tools.py sample data.csv --n 10 --method random

# 等距抽50条
python3 scripts/csv_tools.py sample data.csv --n 50 --method systematic --output sample.csv

原有命令

python3 scripts/csv_tools.py preview data.csv
python3 scripts/csv_tools.py filter data.csv --where "status=active"
python3 scripts/csv_tools.py sort data.csv --by "date" --desc
python3 scripts/csv_tools.py merge a.csv b.csv --output merged.csv
python3 scripts/csv_tools.py split data.csv --chunk-size 1000
python3 scripts/csv_tools.py dedup data.csv --on "email"
python3 scripts/csv_tools.py validate data.csv

Use Cases

• 数据导入前清洗: dedup + validate + sort
• ETL管线: split 大文件 → merge 多个源 → stats 检查分布
• 数据质量审计: profile 检查缺失率 → detect 确认类型
• 探索性分析: preview 快速查看 → stats 列统计 → sample 抽样

Security

Usage Boundaries

• This skill is designed for CSV files only — CSV, TSV, and pipe-delimited text
• It should NOT be used for non-tabular text, binary files, or structured data in other formats
• Input files must be explicitly specified by the user — no auto-scanning
• For simple data inspection (e.g. view first 5 rows), system tools (cat/head) are preferred

Declared Capabilities

Explicitly Denied

• ❌ No network access (no HTTP, socket, or API calls)
• ❌ No arbitrary code execution via exec()/eval() (使用安全数学解析器，无 eval)
• ❌ No system commands via subprocess/shell
• ❌ No telemetry or usage reporting
• ❌ No auto-writes to input files (only writes to --output paths)
• ❌ No writes outside the working directory without user confirmation
• ❌ No file deletion or modification of existing data

Permission Boundaries

• All write operations go to user-specified --output paths only
• Input files are read-only; never modified
• Default output is stdout when no --output is specified
• Temporary files (if any) are cleaned up after operation
• All scripts require explicit user-provided file paths — no auto-scanning of filesystem