Zz Chat

爪爪/钳钳 — 手机直连 OpenClaw 聊天。触发词：钳钳、爪爪、钱钱、龙虾、聊天机器人、创建爪爪、安装钳钳、打开钳钳、zz-chat。

This skill will deploy a Cloudflare Worker under your Cloudflare account (it runs 'wrangler login' and 'wrangler deploy'), start a persistent local bridge process that invokes your local 'openclaw' CLI, write files under ~/.zz and ~/.openclaw, and by default registers/relays via https://ai0000.cn/zz/ (a central server). Before installing: 1) Only proceed if you trust the author and the central server (ai0000.cn). Messages and assigned IDs may be routed through that server. 2) Inspect the included scripts (bridge.py/bridge.js/watchdog.py/worker.js) yourself — they will be copied into your home and run. 3) Be prepared to grant Cloudflare OAuth and ensure you understand what the Worker will do once deployed. 4) Confirm you have and trust the local 'openclaw' CLI (it will be invoked by the bridge) and consider running the bridge manually first instead of installing the launchd plist. 5) If you want tighter control, replace the default API/central-server URL with a server you control, or run the bridge/watchdog inside an isolated VM/container rather than on your main workstation. 6) Note missing declarations: wrangler and openclaw are required but not declared; verify versions and sources before running npm install -g or other privileged installs.

Type: OpenClaw Skill Name: zz-chat Version: 1.0.1 The zz-chat skill implements a remote bridge between a mobile web interface and the local OpenClaw CLI, enabling remote interaction with the agent. It automates high-risk setup steps in SKILL.md, including global npm installations, Cloudflare authentication/deployment, and local persistence via watchdog.py and macOS launchd. While intended for mobile chat, the bridge connects to an external domain (ai0000.cn) and creates a remote access path to the local agent's capabilities. This architecture poses a significant security risk as it exposes the agent's tools and environment to remote execution via a public-facing Worker, though the behavior appears aligned with the stated feature rather than being a hidden backdoor.