boring-youtube-mining

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

YouTube Mining — World Code Edition

YouTube内容挖掘 — World Code版

You pull recent videos from pre-selected YouTube channels, download their transcripts, and generate content ideas the user can actually use. One pass. No back-and-forth.

你可以从预先选定的YouTube频道提取最新视频，下载其字幕，并生成用户可实际使用的内容创意。一站式完成，无需反复沟通。

Before Starting — Check Dependencies

开始前 — 检查依赖项

Run this silently:

bash

which yt-dlp

If
yt-dlp
is NOT installed, stop and tell the user:

yt-dlp
is required but not installed.
Install it for your platform:
macOS:
brew install yt-dlp
Windows:
winget install yt-dlp
Linux:
pip install yt-dlp
or your package manager
Universal:
pip install yt-dlp
Then run
/boring-youtube-mining
again.

Do not proceed without

yt-dlp

. Stop gracefully.

静默运行以下命令：

bash

which yt-dlp

如果未安装
yt-dlp
，请停止操作并告知用户：

需要
yt-dlp
但未安装。
请根据你的平台进行安装：
macOS：
brew install yt-dlp
Windows：
winget install yt-dlp
Linux：
pip install yt-dlp
或使用你的包管理器
通用方式：
pip install yt-dlp
然后重新运行
/boring-youtube-mining
。

未安装

yt-dlp

时请勿继续操作，需优雅终止流程。

Before Starting — Load World Code (Optional)

开始前 — 加载World Code（可选）

Try to read these files silently:

```
world-code/voice.md
```
— Voice rules (NOT applied to output — this is research, not published content)
```
world-code/conversation.md
```
— Bridge structure (walls, struggles, goblins, treasures)

If conversation.md exists: Use the Bridge to filter and prioritize ideas. Every idea gets mapped to a wall, struggle, or goblin. Ideas that don't connect to any get deprioritized (still listed, just flagged as "no Bridge connection").

If conversation.md doesn't exist: Generate ideas purely from transcript analysis. Don't nag about missing files.

尝试静默读取以下文件：

```
world-code/voice.md
```
— 语音规则（不应用于输出内容，仅用于调研）
```
world-code/conversation.md
```
— Bridge结构（包含壁垒、困境、障碍、机遇）

若conversation.md存在： 使用Bridge过滤并优先排序创意。每个创意都需关联到某个壁垒、困境或障碍。无法关联的创意将被降权（仍会列出，但标记为“无Bridge关联”）。

若conversation.md不存在： 仅基于字幕分析生成创意。无需因文件缺失提示用户。

Step 1: Load Channel List

步骤1：加载频道列表

Read

settings/youtube-channels.md

If the file doesn't exist, create it with starter content and tell the user:

I created
settings/youtube-channels.md
with some example channels. Edit it in Obsidian to add the channels you want to mine, then run
/boring-youtube-mining
again.

Starter content for

settings/youtube-channels.md

markdown

undefined

读取

settings/youtube-channels.md

文件。

若文件不存在，则创建包含初始内容的文件并告知用户：

我已创建
settings/youtube-channels.md
并添加了示例频道。请在Obsidian中编辑该文件，添加你想要挖掘的频道，然后重新运行
/boring-youtube-mining
。

settings/youtube-channels.md

的初始内容：

markdown

undefined

YouTube Channels

Add channels you want to mine for content ideas. Use the @ handle format. Add a short description so Claude knows the context.

@AlexHormozi - Business, offers, scaling
@ChrisDo - Branding, pricing, positioning


Stop after creating the file. Let the user configure it first.

---

添加你想要挖掘内容创意的频道，使用@用户名格式。添加简短描述，以便Claude了解背景信息。

@AlexHormozi - 商业、报价、规模化
@ChrisDo - 品牌、定价、定位


创建文件后停止操作，等待用户完成配置。

---

Step 2: Fetch Recent Videos

步骤2：获取最新视频

For each channel in the list, run:

bash

yt-dlp --flat-playlist --playlist-items 1:5 --print "%(id)s | %(title)s | %(duration_string)s" "https://www.youtube.com/@{handle}/videos" 2>/dev/null

Process channels sequentially (not in parallel) to avoid rate limiting.

Present the results as a numbered list:

undefined

针对列表中的每个频道，运行以下命令：

bash

yt-dlp --flat-playlist --playlist-items 1:5 --print "%(id)s | %(title)s | %(duration_string)s" "https://www.youtube.com/@{handle}/videos" 2>/dev/null

按顺序处理频道（不要并行处理），避免触发频率限制。

将结果以编号列表形式呈现：

undefined

Step 3: Download Transcripts

步骤3：下载字幕

For each selected video, first check if a cached transcript exists:

bash

ls "content/youtube-transcripts/{video_id}.md" 2>/dev/null

If cached: Skip download, read from cache. Tell the user:

Transcript for "{title}" already cached — skipping download.

If not cached: Download the auto-generated English captions:

bash

yt-dlp --write-auto-sub --sub-lang en --sub-format vtt --skip-download -o "content/youtube-transcripts/%(id)s" "https://www.youtube.com/watch?v={video_id}" 2>/dev/null

Then clean the VTT file and save as markdown. Create the directory if needed:

bash

mkdir -p "content/youtube-transcripts"

针对每个选定的视频，首先检查是否存在缓存的字幕：

bash

ls "content/youtube-transcripts/{video_id}.md" 2>/dev/null

若存在缓存： 跳过下载，直接读取缓存内容。告知用户：

视频"{title}"的字幕已缓存 — 跳过下载。

若不存在缓存： 下载自动生成的英文字幕：

bash

yt-dlp --write-auto-sub --sub-lang en --sub-format vtt --skip-download -o "content/youtube-transcripts/%(id)s" "https://www.youtube.com/watch?v={video_id}" 2>/dev/null

然后清理VTT文件并保存为Markdown格式。若需要，创建对应目录：

bash

mkdir -p "content/youtube-transcripts"

Cleaning VTT Captions

清理VTT字幕

The raw VTT file contains timestamps and duplicate lines. Clean it:

Strip all timestamp lines (lines matching
```
XX:XX:XX.XXX --> XX:XX:XX.XXX
```
)
Strip the
```
WEBVTT
```
header and
```
Kind:
```
/
```
Language:
```
metadata lines
Remove duplicate consecutive lines (VTT repeats lines across cue boundaries)
Remove position/alignment tags like
```
<c>
```
,
```
</c>
```
,
```
align:start position:0%
```
Join into paragraphs (group sentences, add line breaks at natural pauses)

Save as

content/youtube-transcripts/{video_id}.md

with frontmatter:

markdown

---
video_id: {video_id}
title: {video title}
channel: {channel handle}
date_fetched: {YYYY-MM-DD}
url: https://www.youtube.com/watch?v={video_id}
---

原始VTT文件包含时间戳和重复行。清理步骤如下：

移除所有时间戳行（匹配
```
XX:XX:XX.XXX --> XX:XX:XX.XXX
```
格式的行）
移除
```
WEBVTT
```
头部以及
```
Kind:
```
/
```
Language:
```
元数据行
移除连续重复的行（VTT会在字幕分段处重复内容）
移除位置/对齐标签，如
```
<c>
```
、
```
</c>
```
、
```
align:start position:0%
```
合并为段落（将句子分组，在自然停顿处添加换行）

将清理后的内容保存为

content/youtube-transcripts/{video_id}.md

，并添加前置元数据：

markdown

---
video_id: {video_id}
title: {video title}
channel: {channel handle}
date_fetched: {YYYY-MM-DD}
url: https://www.youtube.com/watch?v={video_id}
---

{video title}

Channel: {channel handle} URL: https://www.youtube.com/watch?v={video_id}

频道： {channel handle} 链接： https://www.youtube.com/watch?v={video_id}

Transcript

字幕

{cleaned transcript text}


Delete the raw VTT file after conversion.

{清理后的字幕文本}


转换完成后删除原始VTT文件。

Edge Cases

边缘情况处理

No English auto-captions available: Skip the video with a message:

No English captions available for "{title}" — skipping.
Very long transcripts (estimated 10,000+ words): Chunk-summarize before idea generation. Break into ~3,000 word sections, summarize each section's key points, then use the summaries for idea generation. Note in the output that the transcript was summarized.

无英文自动字幕可用： 跳过该视频并提示：

视频"{title}"无英文字幕可用 — 跳过。
超长字幕（预计10000词以上）： 在生成创意前先进行分段总结。将字幕拆分为约3000词的段落，总结每个段落的要点，再基于总结内容生成创意。在输出中注明该字幕已被总结。

Step 4: Generate Ideas

步骤4：生成创意

For each transcript, generate ideas through two lenses:

针对每个字幕，从两个维度生成创意：

Lens 1: Direct Response

维度1：直接回应

How would the user respond to this video's ideas? Generate 2-3 ideas:

Agree & Expand — They made a point you also believe. What's your unique angle on it?
Disagree & Counter — They said something you see differently. What's your take?
Build On — They touched on something but stopped short. Where would you take it?

用户会如何回应该视频的观点？生成2-3个创意：

认同并拓展 — 你认同视频中的某个观点，你的独特视角是什么？
反对并反驳 — 你对视频中的某个观点有不同看法，你的观点是什么？
延伸深化 — 视频提及了某个话题但未展开，你会如何深入探讨？

Lens 2: Gap Ideas

维度2：缺口创意

What's missing? Generate 1-2 ideas:

Topics they mentioned but didn't go deep on
Adjacent topics their audience would care about but weren't covered
The question their video raises but doesn't answer

视频遗漏了什么？生成1-2个创意：

视频提及但未深入的话题
目标受众关心但未被覆盖的相关话题
视频引发但未解答的问题

Per Idea, Include:

每个创意需包含以下字段：

Field	Description
Idea title	Sharp, specific — not "Thoughts on pricing" but "Why Hourly Pricing Kills Solo Businesses"
Angle	Direct Response (agree/disagree/expand) or Gap
Source	Video title + timestamp range if identifiable
Your take (1-2 sentences)	The core argument you'd make
Bridge mapping	Which wall/struggle/goblin this connects to (if conversation.md exists)
Content type	Post, thread, email, essay, video script, carousel
Draft hook	One opening line that would stop the scroll

字段	描述
创意标题	清晰具体 — 不要使用“关于定价的思考”，而要用“按时计费为何会拖垮个体创业者”
角度	直接回应（认同/反对/拓展）或缺口
来源	视频标题 + 可识别的时间范围
你的观点（1-2句话）	你要表达的核心论点
Bridge映射	该创意关联到的壁垒/困境/障碍（若conversation.md存在）
内容类型	帖子、推文、邮件、文章、视频脚本、轮播图
草稿钩子	能吸引用户停留的开场句

Quality Over Quantity

质量优先于数量

3-5 ideas per video. Not 10. Not 20. The best 3-5.
Every idea must pass the "would I actually make this?" test
If an idea is generic ("Content is important"), kill it
Ideas without a clear angle aren't ideas — they're topics. Topics aren't useful.

每个视频生成3-5个创意，不要10个或20个，只保留最优质的3-5个。
每个创意都必须通过“我真的会做这个内容吗？”的测试
若创意过于通用（如“内容很重要”），则舍弃
没有明确角度的不能称为创意，只是话题，话题毫无用处

Bridge-First Filtering (when conversation.md exists)

优先基于Bridge过滤（当conversation.md存在时）

After generating all ideas, sort them:

Strong Bridge connection — directly maps to a wall, struggle, or goblin
Loose Bridge connection — related to the user's world but not a direct map
No Bridge connection — interesting but disconnected from their World Code

Group 3 still gets listed but flagged. The user decides if it's worth pursuing.

生成所有创意后，按以下顺序排序：

强Bridge关联 — 直接映射到某个壁垒、困境或障碍
弱Bridge关联 — 与用户的领域相关但未直接映射
无Bridge关联 — 有趣但与用户的World Code无关

第三类创意仍会列出，但会标记出来。由用户决定是否值得跟进。

Step 5: Save Output

步骤5：保存输出

Write the output file to

content-ideas/youtube-mining-{YYYY-MM-DD}.md

markdown

---
type: youtube-mining
date: {YYYY-MM-DD}
channels_mined: [{list of channels}]
videos_analyzed: {count}
ideas_generated: {count}
---

将输出文件写入

content-ideas/youtube-mining-{YYYY-MM-DD}.md

：

markdown

---
type: youtube-mining
date: {YYYY-MM-DD}
channels_mined: [{挖掘的频道列表}]
videos_analyzed: {数量}
ideas_generated: {数量}
---

YouTube Mining — {YYYY-MM-DD}

YouTube内容挖掘 — {YYYY-MM-DD}

Sources

来源

Video	Channel	Ideas
{title}	{channel}	{count}
...

视频	频道	创意数量
{标题}	{频道}	{数量}
...

Ideas

创意列表

From: "{video title}" (@channel)

来自："{视频标题}" (@频道)

1. {Idea Title}

1. {创意标题}

Angle: {Direct Response — Agree/Disagree/Expand | Gap}
Source: {video title}, ~{timestamp context if available}
Your take: {1-2 sentences}
Bridge: {wall/struggle/goblin or "No direct connection"}
Content type: {post/thread/email/essay/video/carousel}
Draft hook: "{opening line}"

{repeat for each idea}

角度： {直接回应 — 认同/反对/拓展 | 缺口}
来源： {视频标题}，~{时间范围（若有）}
你的观点： {1-2句话}
Bridge映射： {壁垒/困境/障碍或 “无直接关联”}
内容类型： {帖子/推文/邮件/文章/视频/轮播图}
草稿钩子： "{开场句}"

{重复上述格式列出每个创意}

Bridge Summary

Bridge关联总结

Strong Connections

强关联创意

{idea} → {wall/struggle} ...

{创意} → {壁垒/困境} ...

No Connection

无关联创意

{idea} — interesting but not tied to your current World Code ...

{创意} — 有趣但与你的当前World Code无关 ...

Next Steps

后续操作建议

Pick an idea and run
```
/boring-copywriting
```
to draft it
Run
```
/boring-social-content
```
to turn an idea into platform-specific posts
Run
```
/boring-remix
```
to generate multiple angles on your favorite idea

---

选择一个创意，运行
```
/boring-copywriting
```
生成完整内容草稿
运行
```
/boring-social-content
```
将创意转化为平台专属帖子
运行
```
/boring-remix
```
为你喜欢的创意生成多个不同角度

---

Step 6: Wrap Up

步骤6：收尾

After saving, tell the user:

Saved {X} ideas to
content-ideas/youtube-mining-{date}.md
.
{count} ideas with strong Bridge connections, {count} without.
Pick an idea and I can draft it with
/boring-copywriting
, turn it into social posts with
/boring-social-content
, or remix it with
/boring-remix
.

That's it. No recap of the process. No lecture on content strategy.

保存完成后，告知用户：

已将{X}个创意保存至
content-ideas/youtube-mining-{date}.md
。
其中{count}个创意与Bridge强关联，{count}个无关联。
你可以选择一个创意，我将通过
/boring-copywriting
生成草稿，或通过
/boring-social-content
转化为社交帖子，也可以通过
/boring-remix
生成多个角度。

操作到此结束。无需回顾流程，无需讲解内容策略。

Key Principles

核心原则

One pass, full output. Fetch → transcripts → ideas → save. No stopping to ask "should I continue?"
Research output, not published content. No voice applied. No polishing. Raw strategic ideas.
Caching is non-negotiable. Never re-download a transcript. Check the cache first, always.
Bridge-first when available. The World Code connection is what makes this useful instead of just interesting.
Sequential channel processing. Don't hammer YouTube with parallel requests.
3-5 quality ideas per video. Restraint is the skill. Anyone can generate 50 mediocre ideas.
Draft hooks matter. An idea without a hook is just a topic. Topics are cheap.
Don't explain the process. The user doesn't need a play-by-play of "now I'm downloading transcripts." Just do it and show the results.

一站式完成，输出完整结果：获取视频→下载字幕→生成创意→保存。无需中途询问“是否继续？”
调研输出，而非发布内容：不应用语音风格，无需润色，仅输出原始战略创意。
缓存是必须的：绝不重复下载字幕。始终先检查缓存。
有Bridge时优先基于Bridge处理：与World Code的关联是本工具实用而非仅有趣的关键。
按顺序处理频道：不要并行请求YouTube，避免触发限制。
每个视频生成3-5个优质创意：克制是关键。任何人都能生成50个平庸创意。
草稿钩子很重要：没有钩子的创意只是话题，话题毫无价值。
无需解释流程：用户不需要了解“我正在下载字幕”这类细节，直接展示结果即可。

boring-youtube-mining

Original

Translation

YouTube Mining — World Code Edition

YouTube内容挖掘 — World Code版

Before Starting — Check Dependencies

开始前 — 检查依赖项

Before Starting — Load World Code (Optional)

开始前 — 加载World Code（可选）

Step 1: Load Channel List

步骤1：加载频道列表

YouTube Channels

YouTube Channels

Step 2: Fetch Recent Videos

步骤2：获取最新视频

Recent Videos

最新视频

@AlexHormozi

@AlexHormozi

@ChrisDo

@ChrisDo

Step 3: Download Transcripts

步骤3：下载字幕

Cleaning VTT Captions

清理VTT字幕

{video title}

{video title}

Transcript

字幕

Edge Cases

边缘情况处理

Step 4: Generate Ideas

步骤4：生成创意

Lens 1: Direct Response

维度1：直接回应

Lens 2: Gap Ideas

维度2：缺口创意

Per Idea, Include:

每个创意需包含以下字段：

Quality Over Quantity

质量优先于数量

Bridge-First Filtering (when conversation.md exists)

优先基于Bridge过滤（当conversation.md存在时）

Step 5: Save Output

步骤5：保存输出

YouTube Mining — {YYYY-MM-DD}

YouTube内容挖掘 — {YYYY-MM-DD}

Sources

来源

Ideas

创意列表

From: "{video title}" (@channel)

来自："{视频标题}" (@频道)

1. {Idea Title}

1. {创意标题}

Bridge Summary

Bridge关联总结

Strong Connections

强关联创意

No Connection

无关联创意

Next Steps

后续操作建议

Step 6: Wrap Up

步骤6：收尾

Key Principles

核心原则

Related Skills

相关工具