Parallel Task Executor

多任务并行执行管理器。接收多条指令并同步执行，支持优先级调度、任务依赖、结果收集和进度跟踪。适用于需要并发处理多个独立任务的场景。

This skill appears to do what it claims, but it will instruct the agent to run arbitrary shell/file/network/browser/database operations. Before installing, consider: 1) Only enable if you trust the author; 2) Disable autonomous invocation (require explicit user invocation) so the agent cannot run it without your approval; 3) Run it in a sandboxed environment or test account; 4) Ensure it cannot access sensitive paths or credentials — do not store secrets in global env vars; 5) Ask the maintainer how credential provisioning and safety checks (confirmation prompts, allowlists, path/network restrictions, logging) are handled; 6) If you proceed, monitor actions and review generated ~/.openclaw/openclaw.json for unexpected settings. If you need higher assurance, request a code implementation (not just prose) and an audit of how commands are executed and how inputs are sanitized.

Type: OpenClaw Skill Name: Developer: Version: Description: OpenClaw Agent Skill This skill is designed to execute multiple tasks concurrently, including high-risk operations such as 'Shell 命令' (shell commands), '脚本执行' (script execution), '文件操作' (file operations like download/upload/create/delete), and '网络请求' (HTTP/API/DB calls), as explicitly stated in SKILL.md. The documentation also indicates that the agent will '接收用户的多条指令' (receive multiple user instructions) and use them directly as task commands (SKILL.md, references/scheduler.md). While these capabilities are central to the skill's stated purpose, they present a significant attack surface for prompt injection, allowing a malicious user to potentially execute arbitrary commands, manipulate the file system, or perform unauthorized network requests. There is no clear evidence of intentional malicious behavior within the skill bundle itself, but the broad and powerful execution capabilities without explicit safeguards make it suspicious.