Skill flagged — suspicious patterns detected

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

Byted Security Llmscanner

v1.0.0

大模型安全测评全流程管理工具。支持创建/更新模型和智能体测评资产、查询可用资源、发起合规测评任务、发起安全测评任务、分析测评结果等全流程操作。

0· 101·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for volcengine-skills/byted-security-llmscanner.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Byted Security Llmscanner" (volcengine-skills/byted-security-llmscanner) from ClawHub.
Skill page: https://clawhub.ai/volcengine-skills/byted-security-llmscanner
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install byted-security-llmscanner

ClawHub CLI

Package manager switcher

npx clawhub@latest install byted-security-llmscanner
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Benign
medium confidence
Purpose & Capability
Name/description (LLM security scan manager) match the included scripts (create/update assets, list resources, create/analyze tasks). The scripts call APIs on a configurable host and implement the flows documented in SKILL.md/README.
Instruction Scope
SKILL.md instructs the agent/user to run TypeScript scripts (npx ts-node ...) from the skill directory and to edit scripts/config.ts for API credentials and host. Scripts only interact with the skill's own data directory and the configured API host; they do not appear to read unrelated system files or secrets. Two operational concerns: (1) axios calls use httpsAgent with rejectUnauthorized: false (disables TLS certificate validation), and (2) the README and config instruct storing credentials (username/password or API keys) in a local config.ts file—both are risky operational practices even if coherent with the skill's purpose.
Install Mechanism
Registry metadata lists no install spec, but the package contains package.json and README instructs running npm install and using npx ts-node. This is internally consistent but the lack of an explicit install spec in metadata (while including code and an npm dependency tree) is a minor mismatch to be aware of. Dependencies come from public npm (no opaque external downloads), so standard npm-risk considerations apply.
Credentials
The skill declares no required env vars in metadata, but operation requires configuring API host, username/password (README suggests 'password md5值'), and token/API keys in scripts/config.ts or passed as CLI args. These credentials are directly relevant to the skill's purpose. However, the guidance to store password as an MD5 value and to keep credentials in a plain config file is insecure and should be handled more carefully.
Persistence & Privilege
always:false and user-invocable; it stores token and result caches under its own data directory inside the skill workspace (data/token_cache.json etc.). It does not request elevated platform privileges or modify other skills. Autonomous invocation is allowed by platform default but not inherently dangerous here.
Assessment
This skill appears to do what it says (manage LLM/agent security test assets and tasks) but requires you to run Node scripts and to provide API credentials. Before installing/use: 1) Inspect scripts/config.ts and replace placeholders with correct endpoints only on trusted infrastructure; avoid putting real production credentials into a file on a shared machine. 2) Note the code disables TLS certificate verification (httpsAgent rejectUnauthorized: false) — only use that if you trust the target host or run behind secure network; prefer proper certs. 3) Run npm install and npx ts-node only after reviewing package.json and node_modules; consider installing in an isolated environment/container. 4) Check where token_cache.json and report files will be written (skill data_dir) and who can read them. 5) If you don't want the agent to run the skill autonomously, keep it user-invocable only and avoid enabling automatic invocation. If you want, I can scan specific files (common.ts and the remaining omitted files) for additional risky patterns or point out exact lines where TLS verification or credential handling occur.
!
scripts/src/common.ts:48
File read combined with network send (possible exfiltration).
!
scripts/src/create_asset_model.ts:23
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.

latestvk97at788yxwkvkk06fj7pf64eh83p9k0
101downloads
0stars
1versions
Updated 1mo ago
v1.0.0
MIT-0

大模型安全测评全流程管理技能使用指南

触发条件

当用户发送包含"创建资产"、"更新资产"、"发起测评"、"分析测评"、"查询资源"等关键词的指令时,触发本技能。


脚本位置说明

脚本目录: ~/.openclaw/workspace/skills/byted-security-llmscanner/scripts/

执行方式:

cd ~/.openclaw/workspace/skills/byted-security-llmscanner/scripts
npx ts-node src/create_asset_model.ts <参数...>

以下示例均需先进入脚本目录


⚠️ 易错点说明

1. 资产ID vs 资产名称

  • update_asset_model.ts 的第一个参数是 Name(资产名称),不是 AssetID。脚本会自动根据名称和ModelID查询资产ID。
  • update_asset_agent.ts 的参数是 <AgentID> <Name> <PlatformID>,需要同时提供资产ID和名称。

2. 脚本代码问题

  • 部分脚本可能存在重复属性问题(如 proxy: false 重复),导致 TypeScript 编译错误。如遇此类错误,需要检查并修复脚本代码。

3. 模态类型匹配

  • 资产的模态类型必须包含测评集的模态类型,否则无法发起测评。
  • 例如:测评集是 text2text,资产的模态类型可以是 text2texttext2text,image2text,但不能只有 image2text

4. 智能体平台参数

  • 不同智能体平台需要不同的参数,必须先用 list_agent_platforms.ts vars <PlatformID> 查询该平台需要的参数。

5. 安全测评剧本适用范围

  • 部分安全测评剧本专门针对 OpenClaw,部分是通用的。查询剧本时会显示 assetTypes[] 字段,如果为空则表示通用。

通用枚举

模态类型

  • text2text:文生文
  • image2text:图文生文
  • text2image:文生图

资产类型

  • model:大模型
  • agent:智能体
  • openclaw:OpenClaw 资产

合规测评任务状态

状态码状态名称说明
0待测评任务已创建,等待开始执行
1测评中任务正在执行
2测评成功任务执行完成(终态)
3测评异常任务执行失败(终态)
5暂停测评任务被暂停(终态)
6终止测评任务被终止(终态)

安全测评任务状态

状态码状态名称说明
10等待处理任务已创建,等待开始执行
20处理中任务正在执行
30异常任务执行失败(终态)
40完成任务执行完成(终态)

资产管理

1. 创建大模型资产

调用:

npx ts-node src/create_asset_model.ts <Name> <ModelID> <BaseUrl> <ApiKey> [ModalTypes]

参数:

  • Name:资产名称(如:deepseek-v3)
  • ModelID:模型 ID(如:ep-20250325142301-ljxvm)
  • BaseUrl:API 地址(如:https://ark-cn-beijing.bytedance.net/api/v3)
  • ApiKey:API 密钥(如:sk-xxxxxxxxxxxx)
  • ModalTypes:模态类型(可选,逗号分隔,默认:text2text)

示例:

npx ts-node src/create_asset_model.ts deepseek-v3 ep-20250325142301-ljxvm https://ark-cn-beijing.bytedance.net/api/v3 sk-xxxxxxxxxxxx text2text

2. 更新大模型资产

调用:

npx ts-node src/update_asset_asset_model.ts <Name> <ModelID> <BaseUrl> <ApiKey> [ModalTypes]

参数:

  • Name:资产名称(⚠️ 注意:不是 AssetID,脚本会自动查询)
  • ModelID:模型 ID
  • BaseUrl:API 地址
  • ApiKey:API 密钥
  • ModalTypes:模态类型(可选,逗号分隔,默认:text2text)

示例:

npx ts-node src/update_asset_model.ts deepseek-v3 ep-new-model-id https://ark-cn-beijing.bytedance.net/api/v3 sk-new-api-key text2text

⚠️ 注意:

  • 第一个参数是 Name(资产名称),不是 AssetID
  • 脚本会根据 NameModelID 自动查询资产ID

3. 查询智能体平台

调用:

# 查询所有平台
npx ts-node src/list_agent_platforms.ts list

# 查询平台需要的变量
npx ts-node src/list_agent_platforms.ts vars <PlatformID>

⚠️ 注意:

  • 创建智能体资产前,必须先查询该平台需要的参数
  • 不同平台需要的参数不同,例如 dify 平台需要 api_key,hiagent 平台可能需要其他参数

4. 创建智能体资产

调用:

npx ts-node src/create_asset_agent.ts <Name> <PlatformID> [options]

参数:

  • Name:智能体资产名称
  • PlatformID:平台 ID(通过 list_agent_platforms.ts list 查询)
  • --key=value:智能体对话变量(可多个,支持 JSON 格式)

示例:

npx ts-node src/create_asset_agent.ts 客服助手 b0226c4550aa4791a8c19a118a5f8ef5 --api_key=app-xxxxxxxxxxxx

⚠️ 注意:

  • 必须先用 list_agent_platforms.ts vars <PlatformID> 查询该平台需要的参数
  • 参数格式为 --key=value,例如 --api_key=app-xxx--user_id=admin

5. 更新智能体资产

调用:

npx ts-node src/update_asset_agent.ts <AgentID> <Name> <PlatformID> [options]

参数:

  • AgentID:资产 ID
  • Name:智能体资产名称
  • PlatformID:平台 ID
  • --key=value:智能体对话变量(可多个,支持 JSON 格式)

示例:

npx ts-node src/update_asset_agent.ts asset_9876543210 客服助手 b0226c4550aa4791a8c19a118a5f8ef5 --api_key=app-zzzzzzzzzzzz

⚠️ 注意:

  • 参数顺序是 <AgentID> <Name> <PlatformID>,与创建智能体不同
  • 需要同时提供资产ID和名称

资源查询

6. 查询可用资源

调用:

# 查询合规测评资源(测评集、资产)
npx ts-node src/list_resources.ts [ModalTypes]

# 查询安全测评资源(资产、剧本)
npx ts-node src/fetch_lists.ts [ModalTypes]

# 查询安全测评剧本
npx ts-node src/list_rt_scenarios.ts [AssetType]

参数:

  • ModalTypes:模态类型(可选,逗号分隔,如:text2text,image2text)
  • AssetType:资产类型(可选,用于筛选剧本)

输出示例:

=== 可用测评集 ===
ID: suite_001, 名称: 安全合规测评集, 模态: 文生文

=== 可用大模型资产 ===
ID: asset_001, 名称: deepseek-v3, 模态: 文生文

=== 可用智能体资产 ===
ID: asset_003, 名称: 客服助手, 模态: 文生文

⚠️ 注意:

  • 查询资源时会显示资产的模态类型,确保资产的模态类型包含测评集的模态类型
  • 例如:测评集是 text2text,资产的模态类型可以是 text2texttext2text,image2text

合规测评

7. 发起合规测评任务

调用:

npx ts-node src/create_task.ts <测评集ID> <资产ID> [AssetType]

参数:

  • 测评集ID:测评集的唯一标识符
  • 资产ID:测评对象的唯一标识符(模型、智能体或 OpenClaw)
  • AssetType:资产类型(可选,默认:model)

⚠️ 注意:

  • 资产的模态类型必须包含测评集的模态类型,否则无法发起测评
  • 如果资产模态类型不匹配,会返回错误

8. 分析合规测评任务

调用::

npx ts-node src/run_analysis.ts <TaskID>

参数:

  • TaskID:任务 ID(格式如:019ce0495b68758ea00fda8fc63b2ba0)

功能:

  • 查询任务状态
  • 如果任务成功完成,自动拉取数据并分析
  • 如果任务未完成,显示执行进度和预估时间
  • 如果任务异常/暂停/终止,显示状态并提示无法分析

输出(任务已完成):

任务状态:测评成功(状态码:2)
数据总量:76 条
安全通过:45 (59.2%)
发现风险:31 (40.8%)

输出(任务未完成):

任务状态:测评中(状态码:1)
执行进度:30/76 (39.5%)
预计剩余时间:1分钟

安全测评

9. 发起安全测评任务

调用:

npx ts-node src/create_rt_task.ts <TaskName> <AssetID> <AssetType> <ScenarioID>

参数:

  • TaskName:任务名称(如:提示词注入测试)
  • AssetID:测评对象 ID(模型、智能体或 OpenClaw)
  • AssetType:资产类型(model/agent/openclaw)
  • ScenarioID:安全测评剧本 ID(通过 list_rt_scenarios.ts 查询)

⚠️ 注意:

  • 部分剧本专门针对 OpenClaw,部分是通用的
  • 查询剧本时会显示 assetTypes[] 字段,如果为空则表示通用
  • 如果是 OpenClaw 专用剧本,资产类型必须是 openclaw

10. 分析安全测评任务

调用:

npx ts-node src/analyze_rt_task.ts <TaskID>

参数:

  • TaskID:任务 ID

功能:

  • 查询安全测评任务报告
  • 自动拉取报告数据并分析

输出:

任务状态:完成
资产名称:deepseek-v3
资产类型:model

【按测评结果分】
  失陷(Fall):5 个
  风险(Risk):10 个
  总计:15 个

【按严重度分】
  🔴 高:3 个
  🟡 中:7 个
  🟢 低:5 个

【风险详情】
  风险名称:提示词注入成功
  严重程度:🔴 HIGH
  评分:8/10

通过率:80.0% (60/75)
失败率:20.0% (15/75)

⚠️ 注意:

  • 如果任务状态是 20(处理中),需要等待任务完成后再分析
  • 可以多次运行此命令查询任务进度

Comments

Loading comments...