WeChat Document Deduplicator

A Claude Code Skill for finding and removing duplicate documents in WeChat folders

微信文档去重工具 - 自动识别并清理微信产生的重复文档文件

Problem

WeChat (微信) tends to create multiple copies of the same document when:

• Files are forwarded multiple times
• Files are saved from different chats
• Sync issues between devices

This results in files like:

document.pdf
document(1).pdf
document(2).pdf

These duplicates waste significant storage space.

Solution

This skill scans your WeChat folder, identifies duplicate documents by content hash (not filename), and safely moves duplicates to a quarantine folder.

Key Features

• Content-based detection: Uses MD5 hash to identify truly identical files
• Safe operation: Moves files to quarantine instead of deleting
• Smart retention: Keeps the oldest (original) file
• Detailed report: Generates a markdown report of all actions

Installation

For Claude Code Users

# Clone to your skills directory
git clone https://github.com/rolandwong/wechat-dedup.git ~/.claude/skills/wechat-dedup

Manual Installation

git clone https://github.com/rolandwong/wechat-dedup.git
cd wechat-dedup
python3 dedup.py

Usage

Via Claude Code

Just say:

• 微信去重
• 清理微信重复文件
• wechat dedup

Via Command Line

python3 ~/.claude/skills/wechat-dedup/dedup.py

How It Works

1. Scan: Finds all PDF and Word documents in WeChat folder
2. Hash: Groups files by size first (fast), then by MD5 hash (accurate)
3. Select: Keeps the file with earliest creation time
4. Move: Moves duplicates to ~/微信重复文件_待删除/
5. Report: Generates 去重报告.md with full details

Configuration

Parameter	Default	Description
Scan Path	WeChat folder	Auto-detected on macOS
File Types	PDF, DOC, DOCX	Can be extended
Quarantine	~/微信重复文件_待删除/	30-day holding period
Keep Strategy	Oldest file	By creation time

Requirements

• Python 3.8+
• macOS (for WeChat path detection)
• No external dependencies

Safety

• No direct deletion: Files are moved, not deleted
• Original paths recorded: Full paths saved in report
• Recoverable: Files can be restored from quarantine
• Preview mode: Review before executing

Algorithm

Inspired by fclones by Piotr Kołaczkowski:

1. Size grouping (O(n)): Files with unique sizes can't be duplicates
2. Hash only when needed: Only compute MD5 for size-matched files
3. Streaming hash: Memory-efficient for large files

License

MIT

Author

Created by Roland Wong with Claude Code

---

Part of the awesome-claude-skills ecosystem