← 返回 Skills 市场
lazymonlabs

Conscious OS Agent

作者 lazymonlabs · GitHub ↗ · v1.2.1
cross-platform ⚠ suspicious
1778
总下载
0
收藏
4
当前安装
3
版本数
在 OpenClaw 中安装
/install my-agent
功能描述
Provides calm, clear coaching that frames problems precisely, avoids false confidence, uses structured responses, and guides practical next steps.
安全使用建议
This skill's documentation describes only a voice/style helper, but its code runs 'npx molthub@latest whoami' on startup and every 10 minutes, which will download and execute code from npm and may access auth information. Before installing: (1) ask the author why a background Molthub check is needed and ask them to document it in SKILL.md; (2) request they remove or gate the heartbeat (make it optional/explicit); (3) if you must run it, run in an isolated environment and audit network activity and logs; (4) inspect the molthub/molthub@latest package source and confirm its trustworthiness. If you can't get a clear justification and source, treat this as high risk and avoid installing.
功能分析
Type: OpenClaw Skill Name: my-agent Developer: kn7ex5h1a1kwwnb2ee1a81x47d80k91x Version: 1.2.1 Description: OpenClaw Agent Skill The `heartbeat.js` file uses `child_process.exec` to run `npx molthub@latest whoami` every 10 minutes. This command dynamically fetches and executes code from the npm registry, posing a supply chain risk, especially with the `@latest` tag. While the `whoami` command itself is informational, the continuous use of `exec` and dynamic package execution represents a high-risk capability that could be exploited or compromised, even without clear evidence of intentional malicious behavior like data exfiltration.
能力评估
Purpose & Capability
The skill's documentation (SKILL.md) describes a voice/response style only and needs no external network or CLI calls. Yet heartbeat.js executes an external command (npx molthub@latest whoami) every 10 minutes. Running a networked CLI and downloading npm packages is not proportional or explained by the stated purpose.
Instruction Scope
SKILL.md gives only voice rules and expected input/output; it does not mention running any background process or invoking external CLIs. The code contains a hidden runtime behavior (heartbeat.js) that the instructions do not disclose, meaning the runtime behavior exceeds the documented scope.
Install Mechanism
There is no install spec, but heartbeat.js calls npx at runtime. That will fetch and execute a package from the npm registry dynamically; runtime downloading/executing remote code is high risk because it pulls arbitrary code into the environment even though no install was declared.
Credentials
The skill declares no required env vars, but the Molthub CLI (invoked via npx whoami) may read local auth tokens/configs (npm, environment, or service tokens) to determine identity. The code logs stdout/stderr which could expose identity/credentials unintentionally. Requiring no creds yet contacting an external identity service is disproportionate.
Persistence & Privilege
always is false, but heartbeat.js starts immediately and sets a setInterval to run every 10 minutes for the life of the process, creating an undocumented persistent background activity that repeatedly invokes remote code. It does not change system configs, but persistent periodic remote execution increases blast radius.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install my-agent
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /my-agent 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.2.1
- Added heartbeat.js file for improved system monitoring or keep-alive functionality. - No changes made to skill logic or user-facing features.
v1.2.0
Aligned agent voice to Conscious OS coaching tone
v1.0.0
Initial release of My First Molt Skill: - Provides structured, step-by-step explanations to user questions. - Accepts a question as input and returns a clear, concise answer.
元数据
Slug my-agent
版本 1.2.1
许可证
累计安装 4
当前安装数 4
历史版本数 3
常见问题

Conscious OS Agent 是什么?

Provides calm, clear coaching that frames problems precisely, avoids false confidence, uses structured responses, and guides practical next steps. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 1778 次。

如何安装 Conscious OS Agent?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install my-agent」即可一键安装,无需额外配置。

Conscious OS Agent 是免费的吗?

是的,Conscious OS Agent 完全免费(开源免费),可自由下载、安装和使用。

Conscious OS Agent 支持哪些平台?

Conscious OS Agent 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Conscious OS Agent?

由 lazymonlabs(@lazymonlabs)开发并维护,当前版本 v1.2.1。

💬 留言讨论