Task Interrupt Pro

Task Interrupt Skill Pro - 精确停止Agent执行中的子任务进程。 Triggers: "停止猫经理", "停止猫工头", "停止猫财子", "停止猫播播", "停止我的任务", "中断猫经理", "解救猫经理", "agent任务卡住", "进程无响应", "子任务卡死"。 注意：不支...

This skill appears to do what it says: create stop flags and send signals to terminate local subagent processes. Before installing, verify that: (1) your agents use the SESSION_ID naming convention the scripts expect so the correct target is chosen; (2) only trusted agents/users can invoke handle-stop.sh (the script acts on files in /tmp owned by the current user or root); (3) writing an audit log to /tmp is acceptable for your environment (it contains SESSION_ID and reason); and (4) you accept that the script will use kill -9 as a last resort (possible data loss). If you need stricter isolation, run agents under distinct OS users or adapt the PID/flag storage path to a more restricted directory.

Type: OpenClaw Skill Name: task-interrupt-pro Version: 1.0.4 The task-interrupt-pro skill bundle is a utility designed to allow an AI agent to manage and terminate its own sub-processes gracefully. The core logic in scripts like handle-stop.sh and task-template.sh includes significant security hardening, such as strict regex validation for session IDs, protection against symlink attacks on PID and flag files, ownership verification, and process command-line validation before issuing kill signals. The documentation (SKILL.md and instructions.md) is well-aligned with the stated purpose and contains no evidence of malicious prompt injection or unauthorized data access.