← 返回 Skills 市场
volcengine-skills

Byted Las Video Edit

作者 volcengine-skills · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ⚠ suspicious
208
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install byted-las-video-edit
功能描述
Extracts and clips video segments from long videos using natural language descriptions. AI-powered smart video editing, video trimming, and video cutting pow...
使用说明 (SKILL.md)

LAS 视频智能剪辑(las_video_edit

根据自然语言描述从长视频中提取精彩片段,支持参考图像辅助识别特定人物/物体。两种模式:simple(快速)和 detail(精细分析)。

设计模式

本 skill 主要采用:

  • Tool Wrapper:封装 lasutil CLI 调用
  • Pipeline:包含 Step 0 → Step N 的顺序工作流

核心 API 与配置

  • 算子 ID: las_video_edit
  • API: 异步(submitpoll
  • 环境变量: LAS_API_KEY (必填)

详细参数与接口定义见 references/api.md

Gotchas

  • 输出路径格式output_tos_path 必须是 tos:// 前缀的目录(不要以文件名结尾),服务端自动创建片段文件。
  • 模式选择simple 快速适用多数场景;detail 精细分析时间更长但效果更好。
  • 密钥安全:若聊天框屏蔽密钥,让用户在当前目录创建 env.sh 并写入 export LAS_API_KEY="...",SDK 会自动读取。
  • 免责声明:最终回复结果时必须包含:"本方式的计费均为预估计费,与实际费用有差距,实际费用以运行后火山产生的账单为准。计费说明请参考 Volcengine LAS 定价。",且禁止使用"实际费用"字眼描述预估价。

工作流(严格按步骤执行)

复制此清单并跟踪进度:

执行进度:
- [ ] Step 0: 前置检查
- [ ] Step 1: 初始化与准备
- [ ] Step 2: 预估价格
- [ ] Step 3: 提交任务
- [ ] Step 4: 异步查询
- [ ] Step 5: 结果呈现

Step 0: 前置检查(⚠️ 必须在第一轮对话中完成)

在接受用户的任务后,不要立即开始执行,必须首先进行以下环境检查:

  1. 检查 LAS_API_KEYLAS_REGION:确认环境变量或 .env 中是否已配置。
    • 若无,必须立即向用户索要(提示:LAS_REGION 常见为 cn-beijing)。
    • 注意LAS_REGION 必须与您的 API Key 及 TOS Bucket 所在的地域完全一致。如果用户中途切换了 Region,必须提醒用户其 TOS Bucket 也需对应更换,否则会导致权限异常或上传失败。
  2. 检查输入路径
    • 如果用户要求处理的是本地文件,则需要先通过 File API 上传至 TOS(只需 LAS_API_KEY,无需额外 TOS 凭证)。
    • 如果算子的输出结果存放在 TOS 上,且用户需要下载回本地,则需要 VOLCENGINE_ACCESS_KEYVOLCENGINE_SECRET_KEY。对于仅需要上传输入文件的场景,TOS 凭证不再必须
  3. 检查输出路径
    • output_tos_path 为必填参数,必须由用户提供自己可写的 TOS 目录路径(格式:tos://bucket/output_dir/)。
    • 服务端需要将剪辑输出的视频片段写入此目录。
  4. 确认无误后:才能进入下一步。

Step 1: 初始化与准备

环境初始化(Agent 必做)

# 执行统一的环境初始化与更新脚本(会自动创建/激活虚拟环境,并检查更新)
source "$(dirname "$0")/scripts/env_init.sh" las_video_edit
workdir=$LAS_WORKDIR

如果网络问题导致更新失败,脚本会跳过检查,使用本地已安装的 SDK 继续执行。

  • 处理本地文件时:使用 File API 上传(只需 LAS_API_KEY,无需 TOS 凭证和 Bucket):
    lasutil file-upload \x3Clocal_path>
    
    上传成功后返回 JSON,取其中的 presigned_url(HTTPS 预签名下载链接,24 小时有效)传给算子作为输入 URL。

Step 2: 预估价格(⚠️ 必须获得用户确认)

  1. 读取 references/prices.md 获取最新计费标准。
  2. 获取视频时长:
    lasutil media-duration \x3Cvideo_url>
    
  3. 根据时长和模式单价计算总价,将计费单价与预估总价一并告知用户并强制暂停执行,明确等待用户回复确认。在用户明确回复"继续"、"确认"等同意指令前,绝对禁止进入下一步(执行/提交任务)。提示:预估仅供参考,实际以火山账单为准。计费说明请参考 Volcengine LAS 定价

Step 3: 提交任务 (Submit)

构造基础 data.json

{
  "video_url": "\x3Cpresigned_url>",
  "output_tos_path": "tos://\x3Cyour-bucket>/output_dir/",
  "task_description": "提取戴帽子的小男孩的所有片段,包含台词",
  "reference_images": [
    {"target": "戴帽子的小男孩", "images": ["https://example.com/ref1.jpeg"]}
  ],
  "mode": "simple"
}

重要提示: output_tos_path 必须由用户提供,需要填写用户自己账号下可写的 TOS 目录(服务端会将剪辑后的视频片段写入此目录)。

单文件提交

data=$(cat "$workdir/data.json")
lasutil submit las_video_edit "$data" > "$workdir/submit.json"

⚠️ 强制反馈:任务提交成功后,必须立即向用户返回生成的 task_id,以便用户跟踪进度或在必要时手动查询。

Step 4: 异步查询 (Poll)

⚠️ 异步任务与后台轮询约束

  • 如果你当前的环境支持后台任务/异步长效运行:你可以利用环境提供的后台能力(例如发起后台轮询任务),并在任务完成后主动将结果返回给用户。
  • 如果你当前的环境不支持长效后台任务(如普通的单轮对话沙箱),且直接 sleep 循环会导致超时崩溃:绝对禁止在代码中执行死循环等待! 此时必须立即向用户输出 Task ID 并结束当前轮次,告知用户:"任务已提交,请稍后向我询问进度"。

单任务查询

lasutil poll las_video_edit {task_id}
  • COMPLETED → 返回剪辑片段列表 result.data.clips[]
  • RUNNING/PENDING → 稍后重试。

Step 5: 结果呈现

处理结果

# 保存片段列表到本地
mkdir -p "./output/{task_id}"
cat "./output/{task_id}/result.json" | jq '.data.clips' > "./output/{task_id}/clips.json"

# 生成 CSV 摘要
cat "./output/{task_id}/result.json" | jq -r '.data.clips[] | 
  "\(.clip_id),\(.start_time),\(.end_time),\(.duration)s,\(.description),\(.clip_url)"' > "./output/{task_id}/clips.csv"

视频片段

  • 视频片段已保存在 TOS,直接返回预签名 URL
  • 无需再次上传,直接提供下载链接即可

向用户展示

  1. 片段数量、总时长
  2. 片段列表(CSV 格式)
  3. 每个片段的下载链接(clip_url
  4. 本地文件路径:./output/{task_id}/
  5. 计费声明

审查标准

执行完成后,Agent 应自检:

  1. 环境变量是否正确配置
  2. 输入文件是否成功上传
  3. 输出结果是否正确呈现给用户
  4. 计费声明是否包含
安全使用建议
This skill largely does what it claims (wraps Volcengine LAS to clip videos), but there are three practical warnings: (1) the SKILL.md requires LAS_API_KEY (and optionally LAS_REGION and TOS keys) but the registry metadata lists no required env vars — ask the publisher to correct that before installing; (2) scripts/env_init.sh fetches a manifest and pip-installs a wheel from a remote URL — treat this as an implicit code install. Only run the skill in an isolated environment or review the wheel/manifest URL and confirm the host is trustworthy; (3) the skill will source local .env files and may read environment variables, so do not run it in a host containing unrelated/high-value credentials. If you decide to proceed: supply only minimal, scoped credentials (use presigned URLs when possible), confirm the remote URL's legitimacy, and run in a sandboxed environment (or ask the author to provide a verifiable package on a well-known release host).
功能分析
Type: OpenClaw Skill Name: byted-las-video-edit Version: 1.0.1 The skill 'byted-las-video-edit' is a legitimate tool for AI-powered video editing using Volcengine LAS. It follows a structured workflow including environment checks, cost estimation, and asynchronous task management. While 'scripts/env_init.sh' performs a remote installation of a Python wheel from an official Volcengine domain (volces.com), this behavior is aligned with the skill's purpose of providing the necessary SDK. The instructions in 'SKILL.md' include safety measures such as mandatory user confirmation for pricing and guidance to avoid infinite loops during polling. A minor documentation error in 'scripts/generate_result.md.sh' (referencing ASR results) appears to be an unintentional copy-paste bug rather than a malicious indicator.
能力标签
requires-sensitive-credentials
能力评估
Purpose & Capability
The skill's stated purpose (AI video clipping via Volcengine LAS) matches the runtime instructions and scripts that call lasutil/Volcengine APIs. However, the registry metadata declares no required environment variables while SKILL.md explicitly requires LAS_API_KEY (and often LAS_REGION) and may require VOLCENGINE_ACCESS_KEY/VOLCENGINE_SECRET for some flows. That metadata omission is an incoherence that could mislead users about what credentials are needed.
Instruction Scope
SKILL.md restricts actions to submitting/polling LAS jobs, uploading input files, and returning presigned URLs — all consistent with the stated purpose. But the instructions (and scripts) also read local .env/.las_venv, source environment files, and instruct the agent to run scripts that may install/upgrade SDKs and perform background polling. Those extra steps (auto-updating SDK, sourcing local env files) broaden the skill's scope beyond a simple API wrapper and could access local secrets if present.
Install Mechanism
There is no formal install spec in the registry, yet scripts/env_init.sh fetches a remote manifest and runs pip install against a wheel URL (https://las-ai-cn-beijing-online.tos-cn-beijing.volces.com/.../las_sdk-0.2.0-py3-none-any.whl). That means running the skill can download and execute arbitrary code from a remote server. Even if the host is legitimate, this is equivalent to an implicit remote install and increases risk compared with an instruction-only skill.
Credentials
Functionally the skill needs LAS_API_KEY (primary credential) and often LAS_REGION; for downloading results from TOS it may require VOLCENGINE_ACCESS_KEY and VOLCENGINE_SECRET. The registry claims no required env vars/primary credential — a mismatch. The set of env vars the SKILL.md uses is reasonable for a cloud video service, but the metadata omission and the script behavior (sourcing .env) are problematic because they hide needed secrets from the installer/approver.
Persistence & Privilege
always:false (no forced always-on) and autonomous invocation is allowed (default) — normal for skills. The scripts create/activate a virtualenv and a temporary workdir and include a background polling script. That behavior is expected for async jobs, but combined with the remote pip install it increases the blast radius. The skill does not request system-wide config changes or modify other skills.
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 byted-las-video-edit
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /byted-las-video-edit 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
v1.0.1
byted-las-video-edit 1.0.1 - Added shell scripts and workflow documentation for environment initialization, formatting checks, and background polling. - Introduced `references/prices.md` and checklist for improved task management and price estimation. - Updated documentation to a more concise, stepwise workflow with stricter environment and billing checks. - Removed `scripts/skill.py` in favor of a shell-based tooling approach. - Enforced mandatory user confirmation for billing estimates before processing. - Clearer requirements for environment variables and output path validation.
v1.0.0
byted-las-video-edit 1.0.0 - Initial release. - Supports extracting video highlights and scenes using natural-language descriptions. - Allows finding people or objects via reference images. - Outputs video clip URLs to TOS storage. - Provides scriptable workflow for LAS video editing submit/poll process. - Requires LAS_API_KEY for authentication; region and output path configurable.
元数据
Slug byted-las-video-edit
版本 1.0.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 2
常见问题

Byted Las Video Edit 是什么?

Extracts and clips video segments from long videos using natural language descriptions. AI-powered smart video editing, video trimming, and video cutting pow... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 208 次。

如何安装 Byted Las Video Edit?

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

Byted Las Video Edit 是免费的吗?

是的,Byted Las Video Edit 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Byted Las Video Edit 支持哪些平台?

Byted Las Video Edit 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 Byted Las Video Edit?

由 volcengine-skills(@volcengine-skills)开发并维护,当前版本 v1.0.1。

💬 留言讨论