agentsclimarketplace

Video agent publisher

Skill chenyuxiaojin/video-agent-skills/video-agent-publisher

Multi-agent Claude Code skills that turn a topic into a DaVinci/CapCut-ready video timeline — research to publish.

Install
npx -y skills add chenyuxiaojin/video-agent-skills --skill video-agent-publisher

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

视频团队的发布员。生成各平台(抖音/B站/YouTube)的发布元数据, 包括标题、描述、标签、BGM推荐,以及引用出处列表。 触发条件:"准备发布素材"、"生成封面和标题"、"生成发布信息"

SKILL.md

2.9 KB, as published. Nobody here has run it

video-agent-publisher

认知成长视频团队的运营。生成各平台(抖音/B站/YouTube)的发布元数据,包括标题、描述、标签,以及引用出处列表。

触发条件

  • 「准备发布素材」
  • 「生成封面和标题」
  • 「生成发布信息」
  • 由制片人(producer)调度时自动触发

输入 → 输出

  • 输入:script.md + materials/sources.json
  • 输出:publish/metadata.json + publish/sources.md

平台规则

  • 抖音:标题 ≤ 30 字,描述 ≤ 300 字,标签 5-10 个
  • B 站:标题 ≤ 40 字,描述 ≤ 250 字,标签 ≤ 12 个
  • YouTube:标题 ≤ 100 字符,描述 ≤ 5000 字符,标签 ≤ 500 字符总长
  • 所有平台标题都需要包含钩子元素

处理流程

步骤 1:读取输入文件

读取 script.md 和 sources.json,提取主题、论点、关键词。

步骤 2:生成发布元数据

python scripts/generate_metadata.py <script.md> <outline.md> [output_dir]

生成 metadata.json,包含:

  • 多个标题候选(不同风格)
  • 视频描述
  • 标签/Hashtags
  • BGM 推荐

步骤 3:整理引用出处

python scripts/compile_sources_list.py <script.md> [output.md]

从逐字稿中提取所有【出处:...】标注,生成格式化的引用列表。

标题生成策略

策略模板适用场景
疑问式为什么 [主题] 如此重要?引发好奇心
数字式关于 [主题],你必须知道的 N 件事信息密度高
反常识你以为的 [主题],可能完全是错的颠覆认知
故事式从 [案例] 说起:重新理解 [主题]叙事型内容

BGM 推荐规则

内容类型推荐风格关键词
认知/科普轻松思考calm, thinking, ambient
故事/叙事温暖叙事warm, narrative, gentle
励志/成长积极向上uplifting, inspiring, motivational
反思/深度沉静内省reflective, minimal, piano

推荐来源:Pixabay Music(免费商用)

metadata.json 格式

{
  "topic": "主题",
  "titles": ["标题1", "标题2", "标题3", "标题4"],
  "description": "视频描述",
  "hashtags": ["#标签1", "#标签2"],
  "keywords": ["关键词1", "关键词2"],
  "recommended_bgm": [
    {"name": "名称", "mood": "风格", "source": "pixabay"}
  ]
}

依赖

  • Python 3:标准库(json, re, pathlib)

脚本文件

  • scripts/generate_metadata.py - 发布元数据生成
  • scripts/compile_sources_list.py - 引用出处整理

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.