其他技能
reverse-engineer-channel
Use when the user wants to reverse engineer how a YouTube channel grew, from its turning points, formats, thumbnails and hooks, into moves to copy and moves to skip. Requires the TubeLab MCP.
tubelab · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add TubeLabHQ/youtube-research-skills --skill reverse-engineer-channel技能说明
name: reverse-engineer-channel description: Use when the user wants to reverse engineer how a YouTube channel grew, from its turning points, formats, thumbnails and hooks, into moves to copy and moves to skip. Requires the TubeLab MCP. license: MIT metadata: author: tubelab version: "5.0" requires: tubelab-mcp
Reverse Engineer Channel
Overview
How one channel grew, worked out from its whole video list and turned into moves the user can copy. Growth comes in steps. A step is a stretch where typical views jump and stay up, after something changed: a format, a thumbnail look, a topic, a length, or a hit the channel then repeated. A trait the best and the worst videos share explains nothing. Every read here is free. Not for diagnosing the user's own channel (channel-analysis).
Workflow
- Clarify scope only if needed. If the user names their own channel or market, aim the moves at it.
- Read the channel's full long-form list and its Shorts. Say how far back the list reaches, and whether the early videos are missing.
- Group the videos by quarter, or by month for a young channel, with the uploads and median views of each period. Find the steps, where the median jumps and stays up.
- For each step, find what changed just before it: the format, topic, title pattern, thumbnail look, length or upload pace. Keep the two or three clearest turning points.
- Group the videos into formats. A format is a title pattern with its slots marked, a thumbnail layout and a usual length. Count how many videos in each beat the channel's average. Note the formats the channel dropped.
- View every thumbnail on the list, on numbered contact sheets. Say so if one fails to load.
- Compare the ten best videos with the ten worst, by ratio. Keep only the differences.
- Read the first 30 seconds of the five best videos and of two weak ones.
- Read the best videos' descriptions for what the channel sells and who sponsors it.
- Write five moves to copy, best first. Each names the skill that runs it, such as
title-formulas,thumbnail-concepts,script-breakdown,niche-bendingormonetization-research, and a prompt the user can paste. Add two or three moves to skip.
Output Format
# 🔬 Reverse Engineering: [{channel}]({url})
- **How it grew:** {one sentence, with numbers}
- **Sample:** {n} long videos from {month year} to {month year} · {n} thumbnails viewed · {n} openings read
- **Shorts:** {whether the channel uses them, and how they do next to the long videos}
- **Start with:** {the one move to copy this week}
## 📈 Growth Timeline
| Period | Uploads | Median views | What changed |
|---|---|---|---|
| {quarter or month} | {n} | {views} | {a few words, or "nothing new"} |
## 🚀 Turning Points
### 1. {What changed} · {month year}
[]({video url})
- [{title}]({video url}) · {views, ratio, date}
- **Before:** {what the channel did until then, with its median views}
- **After:** {what it did next, with its median views}
## 🧩 Formats
| Format | Videos | Beat the average | Best video |
|---|---|---|---|
| "{title pattern with slots}" | {n} | {n} of {n} | [{title}]({video url}) · {views, ratio} |
- **Dropped:** {a format the channel stopped making, and how it did}
## 🔍 Best 10 Against Worst 10
**Best:**
[]({video url}) []({video url}) []({video url})
**Worst:**
[]({video url}) []({video url}) []({video url})
| | Best 10 | Worst 10 |
|---|---|---|
| Title | {what they share} | {what they share} |
| Thumbnail | {what they share} | {what they share} |
| Topic | {what they share} | {what they share} |
| Length | {usual length} | {usual length} |
## 🎣 Hooks
- **What the strong openings share:** {one sentence}
- [{title}]({video url}) · {views, ratio}: "{the first lines, word for word}"
## 💰 How It Earns
- **{Method}:** [{offer or sponsor}]({url}) · seen in [{title}]({video url})
## ✅ Moves to Copy
1. **{Move}**
- **Proof:** [{title}]({video url}) · {views, ratio, date}
- **Run next:** `{skill}` · "{a prompt the user can paste}"
## ❌ Moves to Skip
1. **{Move}:** {why it will not carry over, with proof}
- Length: about 1,200 words. At most ten timeline rows, so merge quiet periods. Two or three turning points, up to five formats, three hooks, five moves to copy.
- Language: plain words and short, complete sentences. Nothing compressed.
- Bullets: one fact each, under 25 words, except quotes, verbatim titles and prompts. No extra sections or bullets.
- If it runs long, cut whole items. Never cut proof, and never squeeze sentences.
- Links: link every video, channel, moment, brand and offer you cite, built from IDs, handles and links in the data. Never guess one.
- Thumbnails: the 320px
mqdefault.jpgimage, linked to its video, on its own line under a heading or up to three on one line under Best and Worst.
Common Pitfalls
- Do not read today's views as what a video got at launch. Old videos have had years to collect views, and last month's uploads are still collecting.
- Do not call one hit a turning point. It counts only if the channel repeated it and the views stayed up.
- Do not state a cause as fact. The data shows what changed before a step, not why viewers came.
- Do not study only the hits. A trait the worst videos also have is not why the best ones won.
- Do not tell the user to copy what only works at this channel's size, such as a famous face, a big budget or rare access. Those are moves to skip.
- Do not mix Shorts into the long-form numbers.