📄 中文摘要
在树莓派上运行的 AI 代理可以管理智能家居仪表板、撰写博客文章、追踪拼车、监控备份等，且大多数操作无需用户事先批准。然而，每次 OpenClaw 更新后，代理会重新开始请求批准，执行命令受到限制，子代理无法生成，导致自主性下降。为了解决这一问题，需要对配置进行调整，并清楚了解所需的安全权衡。
保持您的 OpenClaw AI 代理自主性：配置调整、安全权衡及何时拉紧缰绳
📄 English Summary
Keeping Your OpenClaw AI Agent Autonomous: Config Tweaks, Security Trade-offs, and When to Pull the Leash
An AI agent running on a Raspberry Pi can manage a smart home dashboard, write blog posts, track carpools, and monitor backups, often without prior user approval. However, after each OpenClaw update, the agent tends to revert to requesting approvals, with execution commands being gated and sub-agents failing to spawn, leading to a regression in autonomy. Addressing this issue requires configuration tweaks while being aware of the security trade-offs involved.