WeChat Auto Reply (Cross-Platform v5)

Monitor WeChat for new messages from any specified contact and auto-reply. Supports macOS (Peekaboo CLI) and Windows (PeekabooWin).

Prerequisites

macOS (Apple Silicon / Intel)

1. Peekaboo CLI installed:

   brew install steipete/peekaboo/peekaboo
   

2. Permissions granted: Screen Recording + Accessibility

   peekaboo permissions
   

3. WeChat running and visible on screen

Windows (10 / 11)

1. Node.js 22+ installed: https://nodejs.org

   node --version   # 确认 >= v22.0.0
   

2. PeekabooWin installed:

   git clone https://github.com/FelixKruger/PeekabooWin
   cd PeekabooWin
   npm install
   npm test
   

3. Set environment variable (告诉脚本 PeekabooWin 在哪里):

   # 临时设置（当前 PowerShell 窗口有效）
   $env:PEEKABOO_WIN_DIR = "C:\Users\\x3C你的用户名>\PeekabooWin"
   
   # 永久设置（推荐）
   [System.Environment]::SetEnvironmentVariable("PEEKABOO_WIN_DIR", "C:\Users\\x3C你的用户名>\PeekabooWin", "User")
   

4. WeChat (微信) running and visible on screen (不能最小化)
5. Windows permissions:
   • 部分操作可能需要以管理员身份运行 PowerShell
   • 确保微信窗口未被"专注助手"屏蔽

Python

• Python 3.7+ (macOS 和 Windows 均需要)
• 脚本仅使用标准库，无需额外 pip 安装

Quick Start

Step 0: Platform Detection & Prerequisite Check

When the user invokes this skill for the first time, always run the prerequisite check first. Do NOT start monitoring until all prerequisites are confirmed.

Run these checks using Bash tool:

# 1. Detect platform
echo "PLATFORM: $(uname -s 2>/dev/null || echo Windows)"

# 2. Check Python
python3 --version 2>/dev/null || python --version 2>/dev/null

If macOS (Darwin):

# Check Peekaboo CLI
which peekaboo && peekaboo version

# Check permissions
peekaboo permissions

If Peekaboo NOT found, show this message to the user:

Peekaboo CLI 未安装，请先执行以下命令：

brew install steipete/peekaboo/peekaboo

安装后还需要授予权限：

1. 打开 系统设置 → 隐私与安全 → 屏幕录制 → 勾选 Peekaboo
2. 打开 系统设置 → 隐私与安全 → 辅助功能 → 勾选 Peekaboo

安装完成后对我说"监控 XX 的微信"即可。

If permissions NOT granted, show:

权限未完全授予，请打开以下设置：

1. 系统设置 → 隐私与安全 → 屏幕录制 → 勾选 Peekaboo
2. 系统设置 → 隐私与安全 → 辅助功能 → 勾选 Peekaboo

授权后重新运行 peekaboo permissions 确认。

If Windows:

# Check Node.js version (must be >= 22)
node --version

# Check PEEKABOO_WIN_DIR environment variable
echo $env:PEEKABOO_WIN_DIR

# Check PeekabooWin exists
if (Test-Path "$env:PEEKABOO_WIN_DIR\bin\peekaboo-win.js") { echo "PeekabooWin: FOUND" } else { echo "PeekabooWin: NOT FOUND" }

# Verify PeekabooWin works
node "$env:PEEKABOO_WIN_DIR\bin\peekaboo-win.js" --help 2>$null

If Node.js NOT found or version \x3C 22, show this message:

Node.js 未安装或版本过低，请按以下步骤操作：

第 1 步：安装 Node.js 22+

1. 打开 https://nodejs.org
2. 下载 22.x.x LTS 版本（不要下载 20.x 或更低）
3. 运行安装程序，一路点击"Next"完成安装
4. 安装完成后，重新打开 PowerShell，运行 node --version 确认版本 >= v22.0.0

安装完成后对我说"监控 XX 的微信"即可。

If Node.js OK but PeekabooWin NOT found, show this message:

PeekabooWin 未安装，请按以下步骤操作：

第 2 步：安装 PeekabooWin

1. 打开 PowerShell，运行：

   cd C:\
   git clone https://github.com/FelixKruger/PeekabooWin
   cd PeekabooWin
   npm install
   npm test
   

2. 等待 npm test 全部通过（表示安装成功）

第 3 步：设置环境变量

[System.Environment]::SetEnvironmentVariable("PEEKABOO_WIN_DIR", "C:\PeekabooWin", "User")

设置后必须重新打开 PowerShell 才能生效。

全部完成后对我说"监控 XX 的微信"即可。

If PeekabooWin found but node --help fails, show:

PeekabooWin 已安装但运行异常，请检查：

1. Node.js 版本是否 >= 22：node --version
2. 在 PeekabooWin 目录重新安装依赖：cd C:\PeekabooWin && npm install
3. 确认没有安全软件拦截

All Prerequisites Met

Once all checks pass, proceed to Step 1 below.

Step 1: Ask User for Contact & Style

1. Ask user: "Which contact to monitor?" (required - no default)
2. Ask user: "Reply style/tone?" (default: friendly, match user personality)
3. Take screenshot using the appropriate engine
4. Read screenshot to identify current chat state
5. Start monitor script with --contact parameter (background)
6. Set up automation for auto-reply polling

Example: User says "监控 mini 的微信"

macOS:

• Contact = "mini"
• Start monitor: python3 scripts/wechat_monitor.py --contact mini --interval 600
• Screenshot: peekaboo image --mode window --app 微信 --retina --path /tmp/wechat_mon/latest.png

Windows:

• Contact = "mini"
• Start monitor: python scripts/wechat_monitor.py --contact mini --interval 600
• Screenshot: node \x3CPEEKABOO_WIN_DIR>/bin/peekaboo-win.js screen capture --output \x3CTEMP>\wechat_mon\latest.png

Core Workflow

Step 1: Check Contact & Current State

macOS:

peekaboo list apps --json | grep 微信
peekaboo image --mode window --app 微信 --retina --path /tmp/wechat_mon/latest.png

Windows:

node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js app list
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js screen capture --output %TEMP%\wechat_mon\latest.png

Read the screenshot to:

• Identify the current chat contact
• Check for unread messages (red badge on avatar)
• Note existing messages in the conversation

Step 2: Reply to a Message

macOS (Peekaboo CLI)

# 1. Switch to WeChat
peekaboo app switch --to 微信

# 2. Click input box (coords may vary by window position)
peekaboo click --coords 900,780 --app 微信

# 3. Type the message
peekaboo type "你的回复内容" --app 微信

# 4. Send using type --return (NOT press --key return)
peekaboo type "" --app 微信 --return

IMPORTANT: peekaboo press --key return does NOT reliably trigger WeChat send. Always use peekaboo type "" --app 微信 --return to send.

Windows (PeekabooWin)

# 1. Switch to WeChat
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js app switch --name "微信"

# 2. Click input box (use label or coordinates)
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js click --on "100,200"
# 或者用 label 点击:
# node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js click --on "消息输入框" --snapshot latest

# 3. Type the message
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js type --text "你的回复内容"

# 4. Send using press Enter
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js press --keys "Enter"

Windows 注意: PeekabooWin 支持 --on "label" 模式通过标签点击 UI 元素，比坐标点击更可靠。 建议先用 peekaboo-win see --mode window --title "微信" 确认可点击的元素标签。

Step 3: Verify Send

macOS:

peekaboo image --mode window --app 微信 --retina --path /tmp/wechat_mon/verify.png

Windows:

node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js screen capture --output %TEMP%\wechat_mon\verify.png

Read the verify screenshot to confirm the message appears as a green bubble.

Step 4: Start Background Monitoring

macOS:

nohup python3 scripts/wechat_monitor.py \
  --contact "联系人名称" \
  --interval 600 \
  --threshold 1000 \
  > /tmp/wechat_mon/monitor.log 2>&1 &

Windows (PowerShell):

Start-Process -NoNewWindow python -ArgumentList "scripts/wechat_monitor.py --contact `"联系人名称`" --interval 600 --threshold 1000" -RedirectStandardOutput "$env:TEMP\wechat_mon\monitor.log" -RedirectStandardError "$env:TEMP\wechat_mon\monitor_err.log"

Important:

• Contact name must match exactly as it appears in WeChat
• Can run multiple monitors for different contacts (use different --workdir and --pending)
• --engine auto (default) will auto-detect macOS vs Windows

Step 5: Set Up Auto-Reply Automation

Create a recurring automation (every 30s) that:

macOS Automation Prompt Template:

检查 /tmp/wechat_pending.txt 是否存在。
IF EXISTS:
  1. 读取 pending 文件获取 CONTACT 和 DETECTED 时间
  2. 如果距离 DETECTED 时间 \x3C 10分钟: 结束（继续等待）
  3. 如果 >= 10分钟: 重新截图，分析是否已回复
     - 已回复(绿色气泡): 删除 pending，不回复
     - 未回复(白色气泡): 生成回复并发送
     发送命令:
       peekaboo app switch --to 微信
       peekaboo click --coords 900,780 --app 微信
       peekaboo type "[回复]" --app 微信
       peekaboo type "" --app 微信 --return
IF NOT EXISTS:
  什么都不做

Windows Automation Prompt Template:

检查 %TEMP%\wechat_pending.txt 是否存在。
IF EXISTS:
  1. 读取 pending 文件获取 CONTACT 和 DETECTED 时间
  2. 如果距离 DETECTED 时间 \x3C 10分钟: 结束（继续等待）
  3. 如果 >= 10分钟: 重新截图，分析是否已回复
     - 已回复(绿色气泡): 删除 pending，不回复
     - 未回复(白色气泡): 生成回复并发送
     发送命令(替换 \x3CPEEKABOO_WIN_DIR>):
       node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js app switch --name "微信"
       node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js type --text "[回复]"
       node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js press --keys "Enter"
IF NOT EXISTS:
  什么都不做

Use automation_update tool with:

• scheduleType: recurring
• rrule: RRULE:FREQ=SECONDLY;INTERVAL=30
• prompt: (the template above, filled with actual contact)

Sending Rules

1. Only reply to the target contact — never reply to other people's messages
2. Match user personality — use the user's real voice, not robotic language
3. Reference current activity truthfully — what is the user actually doing right now?
4. Be warm but natural — no excessive emojis, no corporate tone

Input Box Coordinates / Labels

macOS

The input box coordinates depend on WeChat window position. Default: 900,780

peekaboo see --app 微信 --annotate --path /tmp/wechat_coords.png

Windows

PeekabooWin supports label-based clicking (recommended) or coordinate clicking:

# 推荐: 先查看微信窗口有哪些可点击元素
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js see --mode window --title "微信"

# 然后用 label 点击
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js click --on "消息" --title "微信"

# 或用坐标点击
node \x3CPEEKABOO_WIN_DIR>\bin\peekaboo-win.js mouse click --x 900 --y 780

Platform Differences Summary

Troubleshooting

Stopping

macOS:

kill $(pgrep -f wechat_monitor.py)
automation_update --mode delete --id \x3Cautomation-id>
rm -rf /tmp/wechat_mon /tmp/wechat_pending.txt

Windows (PowerShell):

Stop-Process -Name python -Force -ErrorAction SilentlyContinue
# 删除 automation (在 WorkBuddy 中操作)
Remove-Item -Recurse -Force "$env:TEMP\wechat_mon"
Remove-Item -Force "$env:TEMP\wechat_pending.txt"