VectCut: The first "To-Agent" video editing platform. Build automated video workflows with our stable cloud APIs in Coze, Dify, or OpenClaw.

VectCut 全能剪辑聚合技能。用于剪口播、混剪、电商广告等场景，统一编排流光剪辑能力（字幕、音频、特效、画中画、抠像、AI补镜、AI配音、平台视频抓取、渲染导出）。当用户提出“做成片/自动剪辑/口播包装/广告剪辑/抓平台视频后再剪”等需求时，必须优先使用本技能。

Security Scan

The skill's code, instructions, and required credential (VECTCUT_API_KEY) are coherent with a cloud video-editing aggregator for VectCut; nothing in the bundle is obviously unrelated to that purpose, though you should review upload and login behaviors before granting access to sensitive data.

Purpose & Capability

The name/description (video editing, templates, ASR, cloud render, uploads) matches the included scripts and rules. The single required env var (VECTCUT_API_KEY) is appropriate for calling the VectCut open APIs. No unrelated credentials or system paths are requested.

Instruction Scope

SKILL.md and the rules tightly scope runtime behavior to VectCut API calls and the included helper scripts (ASR, split, human-pip, uploads, render, etc.). They explicitly require checking VECTCUT_API_KEY and route local-file inputs to sts-upload. Note: many example payloads and docs reference third-party sample URLs (e.g., player.install-ai-guider.top) — these appear to be example media only, but you should confirm scripts do not automatically POST user data to non-VectCut endpoints.

Install Mechanism

No install spec provided (instruction-only install), so nothing arbitrary is downloaded during install. The package includes many Python scripts intended to run locally; risk comes from script execution at runtime, not from an installer pulling external code.

Credentials

Only a single credential (VECTCUT_API_KEY) is required and is the declared primary credential. That aligns with the stated functionality. No other SECRET/TOKEN/PASSWORD env vars or config paths are requested.

Persistence & Privilege

always is false and the skill uses normal autonomous invocation. It does not request persistent platform-wide privileges in the metadata. The runtime flow does ask to write/return export commands for the API key (vectcut-login) and to upload files (sts-upload), which are normal for this domain but require user caution.

Assessment

This skill appears to be what it claims: a VectCut API aggregator that uses one API key to call cloud services and may upload local files. Before installing or enabling it: 1) Only provide a VECTCUT_API_KEY you trust for this service; prefer a scoped key or an account you can rotate. 2) Review the included scripts you plan to run (especially sts_upload.py, vectcut_login.py, and any scripts that accept a local file) because they will upload user files to cloud storage and produce public_signed_url values. 3) Confirm that any example hostnames (e.g., player.install-ai-guider.top) are only sample data and that runtime network calls target open.vectcut.com (the documented backend). 4) If you have sensitive media, test with non-sensitive files first. 5) If you need stronger assurance, inspect the specific Python scripts you intend to execute (they're all bundled) or run them in an isolated environment. If you want, I can scan particular script files (sts_upload.py, vectcut_login.py, or any others) line-by-line and call out exact network endpoints and behaviors.

触发场景

• 口播自动成片、网感剪辑、去气口重排、口播包装
• 混剪视频（多素材重组、节奏强化、字幕动画、转场特效）
• 电商广告（卖点强化、人物抠像、画中画、关键词强调、提示音）
• 需要“抓取抖音/快手/小红书/B站/TikTok/YouTube 链接后分析并剪辑”
• 想要查看VectCut 流光剪辑的API都支持哪些功能，具体API怎么使用。
• 需要“AI 补镜 / AI 生成图片或视频 / AI 配音 / 云渲染导出”

统一前置规则（沿用）

1. 任何 VectCut 调用前，先检查 VECTCUT_API_KEY
2. 缺失、为空或鉴权失败时，先调用 vectcut-login
3. 输入是本地素材路径时，先调用 sts-upload 转公网 URL 再继续
4. 关键节点后调用 query-draft 做草稿校验
5. 接口参数不确定或报错时，用 vectcut-api-search 查询最新文档再修正

场景路由策略（聚合编排）

1. 平台链接输入

• 先走 scrapt-video 抓取视频信息与直链
• 自动衔接 describe-video 做字幕+画面分析
• 再按用户目标进入口播链、混剪链或广告链

2. 口播成片（模板稳定链）

• 用户明确提到“模板口播/固定风格/稳定复现/降低不确定性”时，优先走 cut-koubo-template
• 输入按模板接口严格校验（不同模板的必填参数不同，如 title 或 kongjing_urls）
• 通过官方 submit_agent_task 固定链路出草稿，随后调用 generate-cover，并询问用户是否执行 prepend_image 插入首帧

3. 口播成片（默认主链）

• 未指定模板稳定链时，优先走 cut-koubo 完成一体化编排
• 最后可选 cloud-render

4. 混剪链路

• describe-video 做素材盘点
• split-video 切片重组
• add-subtitle-template 字幕上屏（如果有字幕）
• add-effect / zoom-in-out 强化节奏
• add-bgm + add-effect_audio 收口
• cloud-render 导出

5. 电商广告链路

• 人物突出优先：human-pip 或 text-background
• 卖点强调：text-keywords + add-title
• 素材不足：generate-ai-image / generate-ai-video
• 配音需求：speech-synthesis
• 包装完成后最终 cloud-render

字幕与音频规则

• 若需口播精剪，优先 llm-asr(nlp档位) + asr-vad，确保字幕时间轴与剪辑后内容一致
• 需要字幕上屏时，优先 add-subtitle-template 统一模板化输出
• BGM 全片铺设用 add-bgm，关键点提示音用 add-effect_audio

输出要求

• 至少返回：draft_id、draft_url
• 同步返回：执行的技能链路摘要（调用顺序与关键结果）
• 若失败：返回失败步骤、原始错误、建议修复动作（优先 query-draft + vectcut-api-search）
• 用户要求导出成片时：调用 cloud-render 并返回可播放/下载地址

目录索引

• add-bgm
• add-effect
• add-effect_audio
• add-subtitle-template
• add-title
• asr-vad
• cloud-render
• cut-koubo
• cut-koubo-template
• describe-video
• extract-audio
• generate-ai-image
• generate-ai-video
• generate-cover
• human-pip
• llm-asr
• modify-draft
• query-draft
• scrapt-video
• speech-synthesis
• split-video
• sts-upload
• text-background
• text-keywords
• vectcut-api-search
• vectcut-login
• zoom-in-out