convez.ai
SkillAutomaçãoAvançado
xjin6

WeChat Rune

xjin6/wechat-rune

Exporte e automatize WeChat com descriptografia local e IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Conjunto de duas skills para Claude Code que permitem exportar histórico de chat do WeChat local (com transcrição de voz via Whisper) e executar um bot de auto-resposta alimentado por IA. Funciona em macOS e Windows, descriptografando dados localmente e usando a chave SQLCipher extraída da memória do processo.

Instalação

$npx convez-ai xjin6/wechat-rune

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 2 meses
Versão
—
Issues abertas
1
Contribuidores
1

Casos de uso

  • Exportar histórico completo de conversas do WeChat com transcrição de áudios
  • Automatizar respostas no WeChat com replies geradas por IA no seu estilo
  • Extrair e analisar dados de chat local sem enviar para servidores terceirizados

Veredito

Projeto sofisticado que tira proveito da integração Claude Code para resolver um problema real de usuários WeChat: recuperar seus próprios dados. O design é engenhoso (extrair chaves da memória), a privacidade é respeitada (tudo local), mas a instalação é complexa e requer conhecimento técnico. Recomendado para quem precisa de controle total sobre dados WeChat.

README

A tradução para o português deste README ainda está na fila.

wechat-rune

Two Claude Code skills for working with your local WeChat (微信) data on macOS and Windows:

SkillPurpose
wechat-rune-exportDecrypt + export chat history to Markdown, with Whisper voice transcription and AI homophone correction
wechat-rune-botClaude-powered auto-reply bot — drafts replies in your voice, sends them through the WeChat UI

Both skills share one key-extraction pipeline and one decryption core. If you've set up the export skill, the bot skill reuses the same scripts/keys/wechat_keys.json and jumps straight to bot config.

Why "rune"

Every WeChat SQLCipher database is sealed behind a 64-character hex key — a string like aa713385968bb2d953fdb6b1f79b83f1e25aac99bd9bb78ea7a31219e274322a. That's a modern rune: a short cryptic symbol carrying the power to unlock everything written behind it. These skills find the runes your WeChat process holds in memory and use them to read your own data back.

Requirements

MacWindows
OSmacOS 13+ (Apple Silicon or Intel)Windows 10+
RuntimePython 3.10+, Xcode CLT (for cc)Python 3.10+
WeChat version4.x (Weixin)4.x (Weixin)
Extra depsbrew install sqlcipher ffmpeg, silk-v3-decoder compiledpip install -r requirements.txt (covers everything)
Privileges neededsudo (once for codesign, once per key extraction)None

Install

Clone, then symlink the two skills into Claude Code's skill directory:

git clone https://github.com/xjin6/wechat-rune.git ~/Desktop/wechat-rune
cd ~/Desktop/wechat-rune
pip install -r requirements.txt

ln -s "$PWD/skills/wechat-rune-export" ~/.claude/skills/wechat-rune-export
ln -s "$PWD/skills/wechat-rune-bot"    ~/.claude/skills/wechat-rune-bot

Relaunch Claude Code. Then:

  • "Export my WeChat chat history" → triggers wechat-rune-export
  • "Set up a WeChat auto-reply bot" → triggers wechat-rune-bot

Repository layout

wechat-rune/
├── skills/
│   ├── wechat-rune-export/
│   │   ├── SKILL.md          # Claude Code skill manifest
│   │   └── README.md
│   └── wechat-rune-bot/
│       ├── SKILL.md
│       └── README.md
├── scripts/                   # shared code used by both skills
│   ├── keys/
│   │   ├── extract_key_macos.c     # Mac: compile once, run with sudo
│   │   └── extract_key_windows.py  # Windows: Python + ctypes
│   ├── export_chat.py
│   ├── transcribe_voices.py
│   ├── start.py                    # bot launcher
│   ├── bot.py                      # bot main loop
│   ├── dashboard.py                # live web dashboard
│   ├── config.py
│   └── core/                       # RAG, embeddings, sender, decrypt, contacts
├── requirements.txt
├── README.md                       # this file
└── README_CN.md                    # 中文版

Privacy & legal

  • Everything runs locally on your own machine against your own decrypted WeChat data.
  • Nothing is uploaded to third-party servers except Anthropic API calls (only when the bot is running, and only the relevant message window, not your full history).
  • The key extraction mechanism (reading your own process memory) is explicitly allowed by macOS/Windows for processes you own. You are not bypassing any server-side security; you are reading the keys your own WeChat client holds for itself.
  • Don't use the bot to impersonate others in contexts where that's misleading or harmful.

License

MIT. See LICENSE.

Credits

Ver o original no GitHub

Bom para

Relacionados