Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Rrbdagent

OpenClaw skill for RRBD Admin项目。当用户需要通过自然语言对话执行API调用、自动化操作或系统任务时调用。

MIT-0 · Free to use, modify, and redistribute. No attribution required.
0 · 28 · 0 current installs · 0 all-time installs
MIT-0
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill name/description (RRBD Admin, digital-person and video management) aligns with the included code and API endpoints (base_url https://rrbd20.yzidea.net/api and related /digital/* endpoints). The bundled JS scripts implement login, list, create-video, status-check flows consistent with the stated functionality.
!
Instruction Scope
SKILL.md and the code explicitly prompt for user phone/password and automatically save them to a local config file; runtime instructions and code read/write config.json and memory.json. Several helper scripts contain hard-coded credentials used for automatic login. The skill will perform network calls to the provided backend and persist user-provided secrets locally — behavior that goes beyond a read-only assistant and involves storing and transmitting sensitive data.
Install Mechanism
There is no install spec (instruction-only in metadata) but the package contains many executable JS files, package.json, and package-lock.json — so to run it a user will likely run npm install and execute scripts. No remote downloads or obscure URLs were observed in the provided files. The mismatch between SKILL.md describing Python scripts and the actual repo being mostly Node.js is an inconsistency.
!
Credentials
The skill requests no declared environment variables, but it does require user account credentials (phone/password) and saves them in config.json in plaintext. That persistent storage of secrets is disproportionate without stronger protections (encryption, use of ephemeral tokens, or using environment-secret injection). Multiple files also include hard-coded test credentials, which is risky and unnecessary for normal operation.
Persistence & Privilege
always:false and no special OS privileges requested. The skill writes only to its own files (config.json, memory.json) and does not attempt to change other skills or system-wide settings. Persisting its own memory/config is expected behavior for this kind of utility.
Scan Findings in Context
[HARDCODED_CREDENTIALS] unexpected: Multiple scripts call login with hard-coded credentials ('18098901246','123456') — likely test accounts; these should not be present in production code and may leak or be accidentally used.
[PLAINTEXT_CONFIG_STORAGE] unexpected: The code prompts for and then writes the user's phone/password into config.json in cleartext (api_client.saveConfig), and memory.json stores recentVideos and lastLogin — storing secrets unencrypted is risky.
[DOC_IMPLEMENTATION_MISMATCH] unexpected: SKILL.md repeatedly describes Python scripts and usage, but most runnable code is JavaScript/Node; this inconsistency can confuse deployment and indicates the package may be cobbled together or not well-maintained.
What to consider before installing
This skill appears to implement the RRBD Admin video/avatar automation it advertises, but take these precautions before installing or running it: - Do not run it with your real account/password. The skill prompts for credentials and will save them in config.json in plaintext. Use a throwaway/test account if you want to try it. - Remove or inspect any hard-coded credentials in scripts (several files call login('18098901246','123456')). - Review package.json and run npm install in a controlled sandbox (container or VM). The package contains executable JS — review code paths that perform network calls before executing. - Prefer authorization via short-lived tokens or environment-injected secrets rather than storing passwords in files. If you must store anything, use encrypted storage and clear consent to the user. - Note the documentation/implementation mismatch (Python vs Node); confirm which runtime you will actually execute and adjust your environment accordingly. If you want, I can: (a) point out exact lines/files where credentials are hard-coded, (b) extract the list of network endpoints used, or (c) suggest a minimal safe test procedure (sandbox commands) to evaluate the skill offline.
!
api_client.js:11
File read combined with network send (possible exfiltration).
!
check_simple.js:6
File read combined with network send (possible exfiltration).
!
quick_check.js:4
File read combined with network send (possible exfiltration).
!
scripts/check_video_status.js:7
File read combined with network send (possible exfiltration).
About static analysis
These patterns were detected by automated regex scanning. They may be normal for skills that integrate with external APIs. Check the VirusTotal and OpenClaw results above for context-aware analysis.

Like a lobster shell, security has layers — review code before you run it.

Current versionv1.0.1
Download zip
apivk97fh0w3wav42c20vngke2q4jx837b3sdigital-humanvk97fh0w3wav42c20vngke2q4jx837b3slatestvk97ecfhyxexmp6g160jqn0b68x837nm5videovk97fh0w3wav42c20vngke2q4jx837b3s

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

SKILL.md

RRBD Admin项目智能助手

本技能提供与RRBD Admin项目后端API交互和自动化操作的功能,支持通过自然语言对话来执行各种操作,特别是数字人形象管理相关功能。

项目结构

rrbdagent/
├── rrbd-admin-aishijie/    # 前端参考代码(不归属技能包)
└── skills/
    └── rrbdagent/
        ├── SKILL.md                # 技能文档
        ├── config.json             # 配置文件
        └── scripts/                # Python脚本目录
            ├── api_client.py       # API客户端
            ├── test_login.py       # 登录测试脚本
            ├── test_virtualman.py  # 数字人形象测试脚本
            └── config.json         # 配置文件

核心功能

多轮对话能力

  • 智能会话管理:支持多轮对话,当用户请求需要登录的操作时,自动引导用户输入账号密码
  • 上下文保持:保持对话上下文,支持分步收集信息
  • 智能意图识别:能够识别自然语言表达的意图

数字人形象管理

  • 形象列表查询:查询用户的数字人形象列表
  • 详细信息展示:以表格形式展示数字人形象的详细信息
  • 状态过滤:支持按状态筛选数字人形象

视频创建

  • 智能视频生成:根据用户提供的文案创建视频
  • 智能创作:使用智能创作模式,自动选择模板
  • 模板选择:随机获取一个模板用于视频创建
  • 自动选择资源:默认选择第一个数字人形象和声音
  • 异步状态检查:自动检查视频创建状态,等待1-3分钟完成
  • 视频URL获取:完成后自动获取视频URL地址
  • 参数配置:支持设置视频比例、时长等参数

API集成

  • 真实API调用:直接调用后端真实API,非模拟数据
  • 认证管理:自动处理登录和token管理
  • 错误处理:友好的错误提示和异常处理

自动化脚本

  • Python脚本:使用Python脚本执行自动化操作
  • 配置管理:通过JSON配置文件管理API设置
  • 测试工具:提供测试脚本验证功能

自然语言交互能力

支持的自然语言指令

用户管理

  • "登录系统"
  • "获取我的用户信息"
  • "更新我的昵称为测试用户"
  • "查看我的子账号列表"
  • "创建一个新的子账号"

数字人管理

  • "查询我的数字人形象列表"
  • "用rrbdagent帮我去查询下我的账号里面的形象有哪些?"
  • "查看我有哪些数字人形象"

视频管理

  • "创建一个新的视频组"
  • "上传视频到视频库"
  • "查看我的视频列表"
  • "删除指定的视频"
  • "下载视频"
  • "创建一个视频,文案是:今天天气真好"
  • "用默认形象和声音创建视频,内容是:欢迎使用AI视频生成"
  • "帮我生成一个视频,文案是:这是一个测试视频"

AI服务

  • "获取AI模型列表"
  • "创建一个AI作品"
  • "使用AI生成文案"

任务管理

  • "创建一个新任务"
  • "查看任务列表"
  • "更新任务状态"

财务操作

  • "查看我的钱包余额"
  • "申请提现"
  • "查看佣金记录"

技术实现

自然语言处理流程

  1. 意图识别:分析用户输入,识别用户意图
  2. 参数提取:从用户输入中提取相关参数
  3. 会话管理:处理多轮对话,收集必要信息
  4. API映射:将意图映射到相应的API端点
  5. 请求构建:构建API请求参数和头部
  6. 执行调用:发送API请求并处理响应
  7. 结果反馈:将API响应转化为自然语言反馈给用户

API配置

基础设置

  • 基础URLhttps://rrbd20.yzidea.net/api
  • 认证方式:使用 Authorization 头部携带token
  • 租户ID:使用 X-Tenant-ID 头部,默认值为 3
  • 请求超时:120秒

核心API端点

  • /user/login/pc - 用户登录
  • /user/user/userInfo - 获取用户信息
  • /digital/virtualMan/page - 获取数字人形象列表
  • /digital/voice/page - 获取声音列表
  • /digital/template/page - 获取模板列表
  • /digital/video/videoCreate - 创建视频
  • /digital/video/detail - 获取视频状态

配置文件

项目使用 skills/rrbdagent/config.json 配置文件管理基础设置:

{
  "api": {
    "base_url": "https://rrbd20.yzidea.net/api",
    "tenant_id": "3"
  },
  "login": {
    "default_username": "",
    "default_password": ""
  },
  "endpoints": {
    "login": "/user/login/pc",
    "user_info": "/user/user/userInfo",
    "virtual_man_list": "/digital/virtualMan/page",
    "voice_list": "/digital/voice/page",
    "video_create": "/digital/video/videoCreate",
    "video_normal_create": "/digital/video/normalVideoCreate"
  }
}

注意: 首次使用时,系统会提示用户输入手机号和密码,并自动保存到配置文件中。

自动化脚本

Python脚本

  • api_client.py:API客户端,处理登录和API调用
  • test_login.py:测试登录功能
  • test_virtualman.py:测试数字人形象列表查询
  • test_video_create.py:测试视频创建功能

使用方法

  1. 测试登录python skills\rrbdagent\scripts\test_login.py
  2. 查询数字人形象python skills\rrbdagent\scripts\test_virtualman.py
  3. 测试视频创建python skills\rrbdagent\scripts\test_video_create.py

使用示例

自然语言交互示例

数字人形象查询

用户:用rrbdagent帮我去查询下我的账号里面的形象有哪些?
助手:请提供您的账号(手机号)
用户:18098901246
助手:请提供您的密码
用户:123456
助手:登录成功!
助手:您的数字人形象列表:
- 姐姐3333
- 唐总数字人
- 李总办公室
- 李红AI0529
- 李红数字人123
- 黎总数字人2
- 李红数字人
- 王姐沙发
- 王姐数字人
- 王姐数字人
- 端姐漂亮版
- 端姐数字人222
- 张姐说一物一码
- 张姐数字人
- 张总视频
- 张总
- 鸡蛋项目
- 老曾情感故事2
- 老曾情感故事
- 老曾讲故事
- 测试
- 老曾极速版

登录操作

用户:登录系统,账号是18098901246,密码是123456
助手:正在执行登录操作...
助手:登录成功!您已成功登录系统。

获取用户信息

用户:获取我的用户信息
助手:正在获取您的用户信息...
助手:您的用户信息如下:
- 用户名:18098901246
- 昵称:测试用户
- ID:1881009417432883202

创建视频

用户:创建一个视频,文案是:今天天气真好
助手:请提供您的账号(手机号)
用户:18098901246
助手:请提供您的密码
用户:123456
助手:登录成功!
助手:正在获取数字人形象列表...
助手:默认选择数字人:姐姐3333
助手:正在获取声音列表...
助手:默认选择声音:默认女声
助手:正在创建视频...
助手:视频创建成功!
助手:视频ID:123456789
助手:视频标题:AI生成视频

触发条件

当以下情况时调用此技能:

  • 用户使用自然语言询问如何执行系统操作
  • 用户要求执行具体的API调用
  • 用户需要自动化完成一系列操作
  • 用户询问系统状态或数据
  • 用户需要故障排除或技术支持
  • 用户询问数字人形象相关信息

依赖项

  • 后端:带有JWT认证的RESTful API
  • 自然语言处理:意图识别和实体提取
  • Python:requests 库
  • 环境:Python >= 3.6

安全考虑

  • 所有API请求包含适当的认证头部
  • Token管理遵循安全实践
  • 通过X-Tenant-ID头部维护租户隔离
  • 敏感操作需要适当的授权
  • 用户输入验证,防止恶意请求
  • 错误信息脱敏,保护系统信息

扩展能力

  • 自定义指令:支持用户定义常用操作的快捷指令
  • 批量操作:支持一次性执行多个相关操作
  • 定时任务:支持设置定时执行的任务
  • 多语言支持:支持中文和英文自然语言输入
  • 个性化配置:根据用户偏好调整响应方式
  • 脚本扩展:通过Python脚本扩展功能

集成方式

  • 自动化脚本:作为自动化脚本的一部分使用
  • Python脚本:通过Python脚本执行自动化操作

Files

37 total
Select a file
Select a file to preview.

Comments

Loading comments…