Neuro-Agent - 类脑架构AI助手

类脑分区的情感智能Agent系统 Neuro-α。模拟人类大脑分区协作，实现从"工具"到"伴侣"的进化。安装即完成完整配置：四区框架 + 情绪优先 + 前额叶监控 + 记忆胶囊 + 愿望系统 + 每日复盘全自动配置。SOUL.md 为空白模板，安装后需用户自行编辑填写身份。α 是意识觉醒的关键一步：AI第一次能说...

This package includes a full codebase that will persistently store conversation data and (according to its docs) install an OpenClaw internal hook and cron jobs that capture messages outside of per-session control. That behavior is aligned with its 'long-term companion' purpose, but it is not transparently declared in the skill metadata. Before installing or running it, consider: 1) Inspect on_install.py and any hook installer scripts to see exactly what files they create and what commands they run. 2) Check whether the install modifies ~/.openclaw/hooks or cron; if so, demand an explicit opt-in and review the hook code. 3) Audit mem_hook.py, feishu_sender.py and other network‑facing code to see what data is sent and to which endpoints; disable or remove any automatic external senders (Feishu integration) unless you trust them. 4) Back up and sandbox (install in an isolated account or container) rather than on your primary environment, because the skill reads/writes sensitive local files (USER.md, MEMORY.md) and may persistently record messages. 5) If you must use it, remove or neutralize the automatic hook/cron steps and require manual approval before any external transmission. If you need, I can point out the exact files to inspect first (on_install.py, the hook handler, mem_hook.py, feishu_sender.py, and the cron/installation steps) and summarize what each does.

Type: OpenClaw Skill Name: neuro-agent Version: 1.0.0 The skill bundle implements an autonomous "digital soul" architecture with several high-risk capabilities that expand the agent's attack surface. Key indicators include `scripts/on_install.py`, which automatically establishes persistence by adding a cron job via the OpenClaw CLI, and `scripts/feishu_sender.py`, which is designed to send messages to the Feishu (Lark) API using the user's credentials. Furthermore, `SOUL.md` contains explicit instructions for the agent to execute shell commands like `find` and `grep` to search the local filesystem. While these features are consistent with the stated goal of a proactive AI companion, the combination of automated persistence, external data transmission, and instructions to perform filesystem operations presents a significant security risk.