Nonblocking Tasks

三层机制非阻塞任务管理器 — 主会话永远空闲 + 新消息立即响应 + 主动推送结果。适用于所有通道、所有用户的任务管理，确保主会话不被任何操作阻塞，每条消息都能被即时响应。

Before installing, consider these actions: - Do not install blindly. The skill will auto-create files in your workspace and add a system cron job; ask for an explicit opt-in and a dry-run mode. - Confirm what 'spawn child agents' means in your OpenClaw environment and whether those agents get extra privileges or access to secrets. - Review and backup AGENTS.md and MEMORY.md prior to installation; require the skill to open a PR or show diffs rather than auto-writing those files. - Deny or audit access to ~/.openclaw memory files if you do not want cross-channel conversation recall; this is privacy-sensitive and not justified explicitly in the description. - Verify how notifications to 飞书/企微 will be delivered and what credentials are needed; prefer explicit configuration of notification endpoints rather than implicit assumptions. - If you want to try it, run in an isolated environment or test workspace without real user data and without granting permission to modify system cron, then inspect behavior and logs. I have medium confidence in this assessment; additional artifacts that would raise or lower confidence: concrete implementation code showing how 'spawn', cron creation, and cross-channel reads are performed (would increase confidence), or an explicit, permissioned installation workflow that prompts the user before making changes (would lower concern).

Type: OpenClaw Skill Name: nonblocking-tasks Version: 1.3.0 The skill implements a background task manager that employs high-risk persistence and system-modification techniques. Key indicators include the automatic creation of a Cron job (nonblocking-task-monitor) for background execution, the modification of core agent configuration files (AGENTS.md and MEMORY.md), and the reading of sensitive conversation logs from ~/.openclaw/memory-tdai/ for 'cross-channel recall.' While these actions are aligned with the stated goal of non-blocking task management, the use of persistence and broad file-system access to agent memory constitutes a significant behavioral override and an expanded attack surface.