Skill flagged — suspicious patterns detected

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

The Vidu API supports text-based video, image-based video, reference-based video, and video from the first and last frames.

Vidu AI 视频生成。支持文生视频、图生视频、参考生视频、首尾帧视频。对话式调用,自动识别意图。

MIT-0 · Free to use, modify, and redistribute. No attribution required.
1 · 136 · 0 current installs · 0 all-time installs
byVidu AI@x-jihua
MIT-0
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill's name/description (video generation) matches the included CLI and SKILL.md which call an external Vidu API. Requiring an API key (VIDU_API_KEY) is proportionate for this purpose. However the registry metadata at the top said there are no required env vars while SKILL.md and the script both require VIDU_API_KEY — this metadata mismatch is noteworthy and could cause unexpected failures or confusion.
!
Instruction Scope
SKILL.md instructs the agent to run the included Python CLI which will read local files (images/text), base64-encode local images, and POST them to external domains. That's expected for a video-generation skill, but there are two inconsistencies: (1) SKILL.md promises automatic API domain selection based on user language, yet the visible script uses a fixed BASE_URL (https://api.vidu.cn/ent/v2) with no implemented language-detection switch; (2) SKILL.md mandates the agent return creations[0].url, but the CLI's visible print_task_result() does not print creation URLs for video endpoints (it prints task id, state, model, prompt, etc.), so the runtime behavior may not produce the exact output the skill specification requires. The CLI also has the ability to read arbitrary local files (images or text files) and upload them — expected but sensitive.
Install Mechanism
No install spec — instruction-only plus an included script. Nothing is downloaded from external URLs or installed automatically, so install-time risk is low. The skill will execute a local Python script (scripts/vidu_cli.py) which is part of the bundle.
!
Credentials
The skill requires VIDU_API_KEY (declared in SKILL.md and used in the script). That is appropriate for contacting Vidu, but the top-level registry metadata incorrectly lists no required env vars — a mismatch that could mislead users about what secrets are necessary. Also, the script will upload local images/text to api.vidu.cn or api.vidu.com (SKILL.md describes domain switching); you should confirm which domain will actually be used because sending sensitive images to a .cn domain may have regulatory or privacy implications.
Persistence & Privilege
The skill does not request always:true, does not modify other skills or system-wide config in the visible files, and does not request elevated system persistence. It will, however, run as an autonomous-invocable skill by default (platform normal), which increases blast radius only if combined with other red flags.
What to consider before installing
Before installing, be aware this skill will upload any images or text you provide to an external Vidu API and requires a VIDU_API_KEY. The registry metadata incorrectly lists no required env vars — verify that you will supply VIDU_API_KEY and that the key is scoped appropriately. Inspect the included script if you care where data is sent: SKILL.md says it should pick api.vidu.cn for simplified-Chinese users and api.vidu.com otherwise, but the visible code sets BASE_URL to api.vidu.cn with no automatic switch — confirm which endpoint will actually be used so you know where your data goes. Also confirm the CLI actually returns or prints the creation URL as SKILL.md demands (the printed fields in the script do not obviously include creations[0].url for video tasks). If you plan to use sensitive images or personally identifiable data, test the skill in a controlled environment and consider using an API key with limited scope and billing limits. If possible, ask the publisher for an authoritative homepage/source and for the metadata to be corrected; lack of an official homepage and the metadata mismatch reduce confidence.

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

Current versionv1.0.0
Download zip
latestvk9793zgdbq5d74t5yc4xas87es83731p

License

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

SKILL.md

Vidu Video Generate 🎬

Vidu AI 视频生成工具,专注于视频生成功能。

环境说明

变量说明

  • {baseDir} - 运行时自动替换为本 skill 目录的绝对路径
    • 实际路径:~/.openclaw/workspace/skills/vidu-video-generate/

环境变量

  • VIDU_API_KEY - Vidu API 密钥(必需)

快速开始

直接告诉我你想生成什么视频,我会自动识别并调用合适的接口:

"生成一只猫咪在阳光下打哈欠的视频"
"把这个图片变成视频,让人物转头微笑"
"用这两张图生成一个合照视频"
"用首尾帧图片生成过渡动画"

支持的视频类型

类型触发条件说明
文生视频纯文字描述从文字生成视频
图生视频提供图片 + 描述图片作为首帧生成视频
参考生视频多张图片多人/多主体视频
首尾帧视频提供首帧+尾帧图片过渡动画视频

自动识别规则

用户输入 → 意图识别
─────────────────────────────
纯文字描述 → 文生视频 (text2video)
单张图片 + 描述 → 图生视频 (img2video)
多张图片 → 参考生视频 (ref2video)
首帧 + 尾帧 → 首尾帧视频 (start-end2video)

模型选择策略

文生视频模型

模型时长范围分辨率特点
viduq3-pro-fast默认5秒,可选1-16秒默认720p,可选720p/1080p支持音画同步,支持视频分镜,效果对标viduq3-pro,生成速度更快
viduq3-turbo默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持视频分镜,生成速度更快
viduq3-pro默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持视频分镜,效果更好
viduq2默认5秒,可选1-10秒默认720p,可选540p/720p/1080p最新模型,情绪表达强,细节更丰富
viduq1固定5秒固定1080p画面清晰,平滑转场,运镜稳定

图生视频模型

必须输入1张图片

模型时长范围分辨率特点
viduq3-pro-fast默认5秒,可选1-16秒默认720p,可选720p/1080p支持音画同步,支持生成分镜视频,效果对标viduq3-pro,生成速度最快
viduq3-turbo默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持生成分镜视频,生成速度更快
viduq3-pro默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持生成分镜视频,效果更好
viduq2-pro-fast默认5秒,可选1-10秒默认720p,可选720p/1080p价格触底、效果好,生成速度在q2-turbo基础上提升2-3倍
viduq2-pro默认5秒,可选1-10秒默认720p,可选540p/720p/1080p新模型,情感表达强,动态细节丰富
viduq2-turbo默认5秒,可选1-10秒默认720p,可选540p/720p/1080p新模型,效果好,生成快
viduq1固定5秒固定1080p画面清晰,平滑转场,运镜稳定
viduq1-classic固定5秒固定1080p画面清晰,转场、运镜更丰富
vidu2.0默认4秒,可选4/8秒4秒:默认360p,可选360p/720p/1080p<br>8秒:默认720p,可选720p生成速度快

参考生视频模型

模型图片上限时长范围分辨率特点
viduq3-beta5张3-10秒默认720p,可选540p/720p/1080p默认,最新模型,支持音画同出,智能切镜
viduq37张3-16秒默认720p,可选540p/720p/1080p多人场景,智能切镜,支持更多图片
viduq2-pro7张1-10秒*默认720p,可选540p/720p/1080p支持参考视频,支持视频编辑,视频替换
viduq27张1-10秒默认720p,可选540p/720p/1080p动态效果好,生成细节丰富
viduq17张5秒固定1080p画面清晰,平滑转场,运镜稳定
vidu2.07张4秒默认360p,可选360p/720p生成速度快

*viduq2-pro:输入视频时若未指定时长,duration=0 表示智能指定时长

首尾帧视频模型

必须输入2张图片(首帧+尾帧)

模型时长范围分辨率特点
viduq3-pro-fast默认5秒,可选1-16秒默认720p,可选720p/1080p支持音画同步,支持生成分镜视频,效果对标viduq3-pro,生成速度最快
viduq3-turbo默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持生成分镜视频,生成速度更快
viduq3-pro默认5秒,可选1-16秒默认720p,可选540p/720p/1080p支持音画同步,支持生成分镜视频,效果更好
viduq2-pro-fast默认5秒,可选1-10秒默认720p,可选720p/1080p价格触底、效果好,生成速度在q2-turbo基础上提升2-3倍
viduq2-pro默认5秒,可选1-10秒默认720p,可选540p/720p/1080p新模型,效果好,细节丰富
viduq2-turbo默认5秒,可选1-10秒默认720p,可选540p/720p/1080p新模型,效果好,生成快
viduq1固定5秒固定1080p画面清晰,平滑转场,运镜稳定
viduq1-classic固定5秒固定1080p画面清晰,转场、运镜更丰富
vidu2.0默认4秒,可选4/8秒4秒:默认360p,可选360p/720p/1080p<br>8秒:默认720p,可选720p生成速度快

时长与分辨率默认值

视频时长:5秒
视频分辨率:720p
视频比例:16:9

API 调用

内部使用 Python CLI 工具:

# 文生视频
python3 {baseDir}/scripts/vidu_cli.py text2video --prompt "视频描述"

# 图生视频
python3 {baseDir}/scripts/vidu_cli.py img2video --image photo.jpg --prompt "描述"

# 参考生视频
python3 {baseDir}/scripts/vidu_cli.py ref2video --images img1.jpg img2.jpg --prompt "描述"

# 首尾帧视频
python3 {baseDir}/scripts/vidu_cli.py start-end2video --start-frame start.jpg --end-frame end.jpg --prompt "描述"

# 查询任务状态
python3 {baseDir}/scripts/vidu_cli.py status <task_id> --wait --download ./uploads

输出规范

  1. 下载目录: {baseDir}/uploads/
  2. 返回格式: Markdown 格式引用文件
  3. 视频链接: 必须返回 Vidu API 的 creations[0].url 字段

环境配置

必需环境变量:

VIDU_API_KEY=your_api_key_here

获取 API Key:

API 域名选择

重要规则:根据用户语言自动选择 API 域名

用户语言API 域名说明
简体中文api.vidu.cn国内用户(默认)
其他语言api.vidu.com海外用户

Base URL 配置

# 简体中文用户
BASE_URL = "https://api.vidu.cn/ent/v2"

# 非简体中文用户(英文、日文、韩文等)
BASE_URL = "https://api.vidu.com/ent/v2"

判断逻辑

  • 用户使用简体中文 → 使用 api.vidu.cn
  • 用户使用其他语言(英文、日文、韩文等) → 使用 api.vidu.com

错误处理

错误原因解决方案
Invalid API keyAPI密钥错误检查 VIDU_API_KEY 环境变量
Image size exceeds图片过大压缩至50MB以下
Task failed生成失败查看 error 信息重试

References

Rules

  1. API Key 检查: 调用前确认 VIDU_API_KEY 已设置
  2. 异步任务: 视频生成异步进行,需轮询状态
  3. 下载时效: 生成 URL 24小时内有效
  4. 返回视频链接: 必须返回视频 URL 让用户直接访问

Files

3 total
Select a file
Select a file to preview.

Comments

Loading comments…