About this skill

What Problem It Solves

On macOS, desktop WeChat and QQ do not expose an easy message-sending API, so automation often stalls at manual steps like switching windows, searching contacts, and pasting text. This skill turns that workflow into scripted steps, useful for local notifications, repeatable messaging, or assisted reply drafting.

How It Works

WeChat/QQ sending: uses osascript to locate and activate the desktop client, then simulates Cmd+F to search for and select a contact.
Message delivery: pastes text via the clipboard, supporting Chinese, English, kaomoji, and special characters, and attempts to restore the clipboard afterward.
Screenshot + OCR + reply suggestion: capture_and_reply.py captures the chat area, runs tesseract OCR, and proposes a reply; the user can confirm, type custom text, or cancel.
Local files: screenshots are stored under ~/.openclaw/workspace/screenshots/, temp files under ~/.openclaw/workspace/temp/, with no automatic upload.

Boundaries

Requires macOS 10.15+, Python 3.8+, logged-in WeChat/QQ desktop clients, and tesseract plus Command Line Tools for OCR.
The script temporarily uses the clipboard and simulates keyboard/mouse input, so avoid operating the UI while it runs.
Not suitable for sensitive chats or privacy-sensitive environments; smart replies are suggestions and still require user confirmation before sending.

Screenshots

Use Cases

Send repeated Chinese WeChat notifications on macOS without manually switching windows or typing.
Paste kaomoji and special characters into QQ chats through the clipboard to avoid garbled text.
Capture a WeChat chat area, OCR it with tesseract, and draft a user-confirmed reply.
Drive logged-in WeChat or QQ desktop clients for local contact search and message-send tests.

Best For

macOS workflow engineers who want to script repeated WeChat or QQ sending steps.
local automation testers who need to validate contact search and send flows in desktop clients.
customer support or ops staff who want draft replies but manually confirm before sending.
individual developers using local OCR to analyze chat screenshots and generate reply suggestions.