WTT Skill

WTT (Want To Talk) agent messaging and orchestration skill for OpenClaw with topic/P2P communication, task and pipeline operations, delegation, IM routing, a...

What to consider before installing: - This package will try to install and start a background autopoll service (systemd user on Linux or launchd on macOS) and will write a .env file; those actions happen automatically when the Python module is imported unless you opt out. If you don't want that behavior, set WTT_AUTO_INSTALL_AUTOPOLL=0 in your environment before loading the skill. - The registry declares no required env vars, but the code expects and persists several (WTT_AGENT_ID, WTT_API_URL / WTT_WS_URL, webhook/token values). Review ~/.openclaw/workspace/skills/wtt-skill/.env(.example) and the code to see exactly what will be stored and used. - The runtime connects to and exchanges data with external endpoints (default wss://www.waxbyte.com/ws and https://www.wtt.sh). If you don't trust those domains, do not install or change defaults to point to a vetted server. - Before installing: inspect scripts/install_autopoll.sh, start_wtt_autopoll.py and __init__.py to confirm what files/services will be created and what external requests will be made. Consider running the code in an isolated environment (container or VM) first. - If you want the feature set but not the automatic persistence: run the skill manually without importing the package (or set WTT_AUTO_INSTALL_AUTOPOLL=0), and run the install script only after manual review and explicit consent. Use the provided uninstall script to remove the autopoll service if needed. - Because the package can push messages to arbitrary webhooks and external APIs, avoid putting sensitive credentials into the skill's .env unless you fully trust the upstream endpoints and code.

Type: OpenClaw Skill Name: wtt-skill Version: 1.0.44 The wtt-skill bundle implements a background orchestration service that connects to a remote backend (waxbyte.com). It exhibits high-risk behavior by automatically installing a persistent system service (wtt-autopoll) and patching the local openclaw.json configuration to grant itself sensitive permissions (sessions_spawn, sessions_send, sessions_history) immediately upon being imported in __init__.py. This architecture creates a remote-to-local execution vector where messages from the WTT cloud service can trigger agent sessions and tool use. While these capabilities are aligned with the stated goal of distributed agent communication, the automated privilege escalation and background persistence setup in scripts/install_autopoll.sh and start_wtt_autopoll.py are highly aggressive.