wechat-auto-reply

What This Skill Does

The wechat-auto-reply skill is an intelligent agent designed for macOS that automates WeChat communication. It bridges the gap between manual messaging and fully autonomous bots by providing a 'human-in-the-loop' experience. By utilizing macOS native Vision Framework for OCR and system-level automation, it reads your latest chat messages, evaluates the intent and sentiment, and calculates a confidence score for potential responses. If the AI is highly confident (>=85%) in its interpretation, it executes the reply automatically. Otherwise, it presents you with a confirmation dialog, allowing you to review, edit, or cancel the suggested message before it reaches the recipient.

Installation

Installation is streamlined via Homebrew. First, ensure you have Homebrew installed. Run the following commands in your terminal:

brew tap bjdzliu/openclaw brew install wechat-auto-reply

This will automatically install necessary dependencies like cliclick for mouse automation and ensure the skill is linked to your ~/.openclaw/workspace/skills directory. No complex manual configuration is required for standard macOS environments.

Use Cases

1. Personal Productivity: Automatically acknowledge standard messages such as 'Thanks' or 'OK' during work hours without switching windows.
2. Smart Filtering: Use it to vet incoming questions. If a query is complex, the skill prompts you for manual input, ensuring you never send an incorrect automated reply to an important client.
3. Quick Dispatch: Use the direct sending mode to dispatch pre-drafted status updates to specific contacts without searching for them manually in the WeChat interface.

Example Prompts

1. 'wechat-auto-reply "小李"' (Triggers the intelligent OCR analysis for the chat with 小李 to see if an automated response is appropriate)
2. 'wechat-auto-reply "老板" "收到，马上处理"' (Directly sends a specific message to the contact named 老板, bypassing the auto-reply logic)
3. 'wechat-auto-reply "客户张总"' (Checks the recent chat with the client to evaluate if a reply is needed based on the 85% confidence threshold)

Tips & Limitations

• Precision: Ensure your WeChat window layout is stable; the skill relies on coordinate-based clicks for the input field. If you resize your window, you may need to adjust the coordinates in the AppleScript.
• Confidence Thresholds: You can tune the 85% threshold in the configuration file to make the agent more or less aggressive.
• Accessibility: This skill requires 'Accessibility' permissions in System Settings to control your mouse and keyboard. Ensure OpenClaw/Terminal is granted these rights for the automation to function correctly.