Cómo usar
- Copia el comando y revisa el paquete antes de instalarlo en tu entorno de agente.
- Consulta las herramientas, credenciales y entornos necesarios en el documento original.
- Describe las entradas y el resultado deseado; contrasta la salida con las fuentes.
npx skills add https://github.com/tubelabhq/youtube-research-skills --skill reverse-engineer-nicheAntes de empezar
- La instalación no incluye suscripción al modelo, cuota API ni permisos sobre tus archivos.
- Las estrellas e instalaciones reflejan actividad, no una prueba independiente de calidad o seguridad.
Contenido del repositorio de origen con títulos, ejemplos, código, tablas, enlaces e imágenes preservados.
Reverse Engineer Niche
Overview
A niche taken apart through 12 to 20 channels, emerging ones first, as a niche guidelines book. Five lenses:
- Working formats: a new video must fit the format and length viewers already watch.
- New and specific: new enough to stop a scroll, specific enough to feel real: a price, a job, a place.
- Anchors: names, brands, tools or terms that tell YouTube which viewers to send.
- Dead versions: a winner's promise that flopped.
- Who can copy it: anyone, or only someone with its asset: experience, trust, credentials, a skill, fame or budget.
Not for last month's winners (trend-report), one channel (reverse-engineer-channel) or whether to enter (niche-validator).
Without TubeLab: no search for new or similar channels, no audience, income, lost-ads or terminated-channel data. The report covers only channels you already know.
Workflow
- Ask once, with any scope question: end with video ideas, for their channel or a new one here?
- Run the searches below, up to 15, never paging one.
- Verify each channel with a free read (it gives the avatar): on topic, long videos, the user's language. English titles can hide Hindi audio; the hit's audience region shows it. Keep 12 to 20.
- Median each channel's last three months of long videos. Winners: its top three at twice that or more. Flops: under half, older than a month, trailing their neighbours. One set of totals everywhere.
- Fill the audience and income page from channel hits only; video hits carry just ads and RPM.
- View every winner's and every flop's thumbnail on contact sheets.
- Group winners into three to five patterns, each on three or more channels, each winner explained against its channel's normal videos.
- Count title formulas and thumbnail compositions across winners and flops. Rising: also in the trending search.
- Read the opening 30 seconds of ten or more winners on different channels. A hook formula is a beat order, named to fit any video, shared by two or more channels.
- Read four to six full winning transcripts from different patterns. Mark each beat and its time. A video formula is a layout two or more share.
- Pair dead versions with their winners, same channel when you can, with the reason.
- Count topics, sub-topics and power words: winning titles against flops. Worn out: as common in flops.
- Run the risk searches and fill the risk page.
- Blend two patterns into a new angle no sampled video has filmed.
- Given a channel, find up to three gaps: patterns its competitors win with that it skips.
- If wanted, run
video-ideationon these findings for the last chapter. Unanswered: offer it in the closing line.
// One search per line. {niche}: one to three phrasings of one niche. {code}: the user's language.
// Keep publishedAtFrom and classificationQuality as written. Only the risk searches leave quality out.
// 10 searches as written. Spend the spare five on a second phrasing, a wider date range, or handle lookups
// ({ "query": ["@handle"] }, nothing else) for big sampled channels found only by video, so the audience page covers them.
// Channels created this year, then emerging channels.
{ "query": ["{niche}"], "language": ["{code}"], "publishedAtFrom": "trimester", "classificationQuality": "neutral", "joinedDateFrom": "year", "medianViewsFrom": 1000 }
{ "query": ["{niche}"], "language": ["{code}"], "publishedAtFrom": "trimester", "classificationQuality": "neutral", "subscribersFrom": 3000, "subscribersTo": 300000, "avgViewsToSubscribersRatioFrom": 1 }
// Outliers: small channels' breakouts, breakouts on channels with 30 videos or fewer, then established channels' breakouts.
{ "query": ["{niche}"], "language": ["{code}"], "type": "video", "publishedAtFrom": "trimester", "classificationQuality": "neutral", "averageViewsRatioFrom": 2, "subscribersCountFrom": 3000, "subscribersCountTo": 300000, "size": 40 }
{ "query": ["{niche}"], "language": ["{code}"], "type": "video", "publishedAtFrom": "trimester", "classificationQuality": "neutral", "averageViewsRatioFrom": 2, "channelVideoCountTo": 30, "subscribersCountFrom": 1000, "size": 40 }
{ "query": ["{niche}"], "language": ["{code}"], "type": "video", "publishedAtFrom": "trimester", "classificationQuality": "neutral", "averageViewsRatioFrom": 1.5, "subscribersCountFrom": 300000, "size": 40 }
// Similar channels, one seed per search: the best breakout, or the given channel. Several seeds blur into a generic set,
// and the hits are mostly small, recently found channels: established ones come from the outlier search above.
{ "relatedChannelId": ["<channelId>"], "language": ["{code}"], "publishedAtFrom": "trimester", "classificationQuality": "neutral", "subscribersFrom": 3000, "size": 40 }
// Outliers of similar channels, same seed.
{ "relatedChannelId": ["<channelId>"], "type": "video", "publishedAtFrom": "trimester", "classificationQuality": "neutral", "averageViewsRatioFrom": 2, "size": 40 }
// Trending title formulas whose use grew 10% or more over the previous period. English only: pass no language.
// Keep only formulas whose example titles fit the niche; the rest are generic.
{ "query": ["{niche}"], "lastVideosCountChangePercentageFrom": 10 }
// Risk, seeded with the three strongest emerging channels (live seeds work): similar channels terminated, then that lost ads.
{ "relatedChannelId": ["<channelId>", "<channelId>", "<channelId>"], "terminated": true, "publishedAtFrom": "year", "size": 40 }
{ "relatedChannelId": ["<channelId>", "<channelId>", "<channelId>"], "monetizationChanged": true, "monetizationAdsense": false, "publishedAtFrom": "year", "size": 40 }
// AI-made channels: count detections.madeWithAI on the channel hits above. A search filtered on it returns AI films, off topic.Output Format
One HTML book from this template, style and script unchanged. Repeat or drop page types as the data allows.
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<title>{Niche}: Niche Guidelines</title>
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Archivo:wght@400;600;700;800&display=swap" rel="stylesheet">
<style>
:root{--paper:#000;--bg:#0e0e0e;--card:#1a1a1a;--fg:#f4f4f4;--mut:#9d9d9d;--line:#2e2e2e;--red:#ff4e42;--pw:min(100vw,177.78vh);--u:calc(var(--pw)/100);color-scheme:dark}
@media(prefers-color-scheme:light){:root{--paper:#e2e2e2;--bg:#fff;--card:#f1f1f1;--fg:#141414;--mut:#666;--line:#e0e0e0;--red:#e3241b;color-scheme:light}}
*{box-sizing:border-box;margin:0;padding:0}
html{background:var(--paper);scroll-snap-type:y mandatory;scrollbar-width:none}
body{background:var(--paper);color:var(--fg);font-family:Archivo,system-ui,sans-serif;display:flex;flex-direction:column;align-items:center;counter-reset:pg;-webkit-print-color-adjust:exact;print-color-adjust:exact}
a{color:inherit;text-decoration:none}a:hover{text-decoration:underline}
.chapter{display:contents}
.page{position:relative;overflow:hidden;width:var(--pw);height:calc(var(--pw)*.5625);margin-block:calc((100vh - var(--pw)*.5625)/2);padding:calc(var(--u)*3.6) calc(var(--u)*5) calc(var(--u)*5);background:var(--bg);display:flex;flex-direction:column;gap:calc(var(--u)*1.8);scroll-snap-align:center;scroll-snap-stop:always;counter-increment:pg;font-size:calc(var(--u)*1.3);line-height:1.4}
.page::before,.page::after{position:absolute;bottom:calc(var(--u)*1.9);font-size:calc(var(--u)*.85);color:var(--mut);letter-spacing:.04em}
.page::before{content:var(--book) " · " var(--ch);left:calc(var(--u)*5)}
.page::after{content:counter(pg);right:calc(var(--u)*5)}
.k{font-size:calc(var(--u)*.95);font-weight:700;letter-spacing:.14em;text-transform:uppercase;color:var(--mut)}
h1{font-size:calc(var(--u)*6.4);font-weight:800;line-height:1;letter-spacing:-.02em}
h2{font-size:calc(var(--u)*2.7);font-weight:800;line-height:1.1;letter-spacing:-.01em}
.lead{font-size:calc(var(--u)*1.9);font-weight:600;line-height:1.3}
.big{font-size:calc(var(--u)*4.6);font-weight:800;line-height:1}
.mut{color:var(--mut)}
small{font-size:calc(var(--u)*.95)}
.f{font-size:calc(var(--u)*3.1);font-weight:800;line-height:1.3;letter-spacing:-.01em}
mark{background:var(--red);color:#fff;padding:0 .25em;border-radius:.25em;-webkit-box-decoration-break:clone;box-decoration-break:clone}
.fill{flex:1;min-height:0;align-content:space-evenly}
.col.fill{justify-content:space-evenly}
.split{flex:1;min-height:0;display:grid;grid-template-columns:1.2fr 1fr;gap:calc(var(--u)*4);align-items:center}
.col{display:flex;flex-direction:column;gap:calc(var(--u)*1.4);min-width:0}
.g2,.g3{display:grid;gap:calc(var(--u)*1.6);align-content:start}
.g2{grid-template-columns:repeat(2,1fr)}.g3{grid-template-columns:repeat(3,1fr)}
.kv{display:grid;grid-template-columns:auto 1fr;gap:calc(var(--u)*.9) calc(var(--u)*2);align-content:start}
.kv dt{color:var(--mut);font-size:calc(var(--u)*1);font-weight:700;letter-spacing:.1em;text-transform:uppercase;padding-top:.3em}
.v{display:flex;flex-direction:column;gap:calc(var(--u)*.4);font-size:calc(var(--u)*1.05);line-height:1.3;min-width:0;background:var(--card);border-radius:calc(var(--u)*.9);padding:calc(var(--u)*.65)}
.v img,.wall img,.anno img{display:block;width:100%;aspect-ratio:16/9;object-fit:cover;border-radius:calc(var(--u)*.6);background:var(--line)}
.v b,.r{align-self:flex-start;display:inline-block;width:fit-content;background:var(--red);color:#fff;font-weight:800;line-height:1;padding:.2em .4em;border-radius:.35em;white-space:nowrap}
.v b{font-size:calc(var(--u)*1.4)}
.r{font-size:.9em}
.x+b,.x+div>b,.no b,.r.lo{background:var(--bg);color:var(--mut);box-shadow:inset 0 0 0 1px var(--line)}
.v span{display:-webkit-box;-webkit-line-clamp:2;-webkit-box-orient:vertical;overflow:hidden}
.v small{color:var(--mut)}
.one .v span{-webkit-line-clamp:1}
.tight .col{gap:calc(var(--u)*.8)}.tight .legend,.tight .kv{font-size:calc(var(--u)*1.15)}.tight .legend{gap:calc(var(--u)*.5)}.tight .kv{gap:calc(var(--u)*.6) calc(var(--u)*2)}.one small{font-size:calc(var(--u)*.85)}
.v.h{display:grid;grid-template-columns:calc(var(--u)*14) 1fr;gap:calc(var(--u)*1);align-items:start}
.v.h>div{display:flex;flex-direction:column;gap:calc(var(--u)*.35)}
.v p{font-size:calc(var(--u)*1.05);line-height:1.35}
.vs{flex:1;min-height:0;display:grid;grid-template-columns:repeat(2,minmax(0,calc(var(--u)*36)));justify-content:space-around;align-content:center;gap:calc(var(--u)*5)}
.vs .v{gap:calc(var(--u)*.8);font-size:calc(var(--u)*1.2);padding:calc(var(--u)*1.4)}
.vs .v>div:first-child{display:flex;justify-content:space-between;align-items:center}
.vs .v b{align-self:center;font-size:calc(var(--u)*3.4)}
.x{position:relative;display:block}
.x::after{content:"";position:absolute;inset:0;border-radius:calc(var(--u)*.6);box-shadow:inset 0 0 0 calc(var(--u)*.22) var(--red);background:linear-gradient(to top right,transparent 49.4%,var(--red) 49.4% 50.6%,transparent 50.6%),linear-gradient(to top left,transparent 49.4%,var(--red) 49.4% 50.6%,transparent 50.6%)}
.anno{position:relative}
.anno i{position:absolute;border:calc(var(--u)*.22) solid var(--red);border-radius:calc(var(--u)*.5)}
.anno i::before{content:attr(data-n);position:absolute;left:calc(var(--u)*-.22);top:calc(var(--u)*-.22);background:var(--red);color:#fff;font:800 calc(var(--u)*1)/1 Archivo,system-ui,sans-serif;padding:calc(var(--u)*.35) calc(var(--u)*.55);border-radius:calc(var(--u)*.4) 0 calc(var(--u)*.4) 0}
.legend{list-style:none;counter-reset:n;display:flex;flex-direction:column;gap:calc(var(--u)*1)}
.legend li{counter-increment:n;display:flex;gap:calc(var(--u)*1);align-items:baseline}
.legend li::before{content:counter(n);flex:none;background:var(--red);color:#fff;font-weight:800;font-size:calc(var(--u)*.9);width:calc(var(--u)*1.8);height:calc(var(--u)*1.8);border-radius:50%;display:grid;place-items:center;transform:translateY(-.1em)}
.beats{display:flex;flex-wrap:wrap;gap:calc(var(--u)*1) calc(var(--u)*3)}
.beats span{position:relative;border:calc(var(--u)*.14) solid var(--line);border-radius:calc(var(--u)*.6);padding:calc(var(--u)*.7) calc(var(--u)*1.2);font-size:calc(var(--u)*1.7);font-weight:700}
.beats span+span::before{content:"→";position:absolute;left:calc(var(--u)*-2.3);color:var(--mut)}
.q{border-left:calc(var(--u)*.2) solid var(--line);padding-left:calc(var(--u)*1.4);display:flex;flex-direction:column;gap:calc(var(--u)*.5)}
.q p{font-size:calc(var(--u)*1.45)}
.q p>b{color:var(--mut);font-size:calc(var(--u)*.85);letter-spacing:.1em;text-transform:uppercase;margin-right:.6em}
.gantt{display:grid;grid-template-columns:auto 1fr;gap:calc(var(--u)*.7) calc(var(--u)*1.6);align-items:center;font-size:calc(var(--u)*1.15)}
.gantt div{position:relative;height:calc(var(--u)*1.5);background:var(--card);border-radius:calc(var(--u)*.3)}
.gantt div i{position:absolute;inset-block:0;left:calc(var(--a)*1%);width:max(calc(var(--u)*.6),calc((var(--b) - var(--a))*1%));background:var(--fg);border-radius:calc(var(--u)*.3)}
.gantt .pay i{background:var(--red)}
.chips{display:flex;flex-wrap:wrap;gap:calc(var(--u)*.8);align-content:start}
.chip{border:calc(var(--u)*.12) solid var(--line);border-radius:calc(var(--u)*.8);padding:calc(var(--u)*.6) calc(var(--u)*1.1);font-size:calc(var(--u)*1.15)}
div.chip>b{background:var(--red);color:#fff;padding:0 .35em;border-radius:.3em;-webkit-box-decoration-break:clone;box-decoration-break:clone}
.chip.worn>b{background:var(--card);color:var(--mut);text-decoration:line-through;box-shadow:inset 0 0 0 1px var(--line)}
.tag{font-size:calc(var(--u)*1.05);font-weight:800;letter-spacing:.14em;text-transform:uppercase}
.no .tag{color:var(--red)}
.t{width:100%;border-collapse:collapse;font-size:calc(var(--u)*1.15)}
.t td,.t th{text-align:left;padding:calc(var(--u)*.65) calc(var(--u)*1.2) calc(var(--u)*.65) 0;border-bottom:calc(var(--u)*.08) solid var(--line);vertical-align:top}
.t th{color:var(--mut);font-size:calc(var(--u)*.85);letter-spacing:.1em;text-transform:uppercase}
.chans{display:grid;grid-template-columns:repeat(4,1fr);gap:calc(var(--u)*.9);align-content:start}
.chans a{background:var(--card);border-radius:calc(var(--u)*.6);padding:calc(var(--u)*.8) calc(var(--u)*1.1);font-size:calc(var(--u)*1);line-height:1.35}
.chans b{display:block;font-size:calc(var(--u)*1.15)}
.chans .av{float:left;margin:calc(var(--u)*.15) calc(var(--u)*.8) 0 0;width:calc(var(--u)*2.6);height:calc(var(--u)*2.6);border-radius:50%;object-fit:cover;background:var(--line)}
.chans .me{box-shadow:inset 0 0 0 calc(var(--u)*.15) var(--fg)}
.cover{justify-content:flex-end}
.cover::before{display:none}
.wall{position:absolute;inset:-12% -6% -12% 22%;display:grid;grid-template-columns:repeat(4,1fr);gap:calc(var(--u)*.8);transform:rotate(-8deg);opacity:.75}
.cover .shade{position:absolute;inset:0;background:linear-gradient(90deg,var(--bg) 34%,transparent 75%),linear-gradient(0deg,var(--bg) 4%,transparent 40%)}
.cover>div:last-child{position:relative;display:flex;flex-direction:column;gap:calc(var(--u)*1.4);max-width:60%}
.divider{justify-content:center}
.divider .n{font-size:calc(var(--u)*13);font-weight:800;line-height:.9;color:transparent;-webkit-text-stroke:calc(var(--u)*.12) var(--mut)}
.divider h1{font-size:calc(var(--u)*7)}
code{font-family:ui-monospace,Menlo,monospace;font-size:.9em;background:var(--card);padding:.1em .4em;border-radius:.3em}
@media screen and (max-width:760px){
:root{--pw:100vw;--u:11px}
html{scroll-snap-type:none}
.page{width:100%;height:auto;min-height:0;margin:0 0 10px;padding:28px 16px 48px}
.page::before{left:16px;bottom:16px}.page::after{right:16px;bottom:16px}
.split,.g2,.g3,.vs{grid-template-columns:1fr}.chans{grid-template-columns:1fr 1fr}.vs{gap:16px}
.cover{min-height:80vh}h1{font-size:42px}.divider h1{font-size:52px}.cover>div:last-child{max-width:none}.wall{inset:-10% -30% 40% -10%}.cover .shade{background:linear-gradient(0deg,var(--bg) 55%,transparent 90%)}
.divider{min-height:60vh}
.beats{padding-left:28px}
}
@page{size:16in 9in;margin:0}
.page{background-image:url("data:image/svg+xml,%3Csvg xmlns=%22http://www.w3.org/2000/svg%22 fill=%22none%22 viewBox=%220 0 22 22%22%3E%3Cmask id=%22tlm%22 width=%2222%22 height=%2222%22 x=%220%22 y=%220%22 maskUnits=%22userSpaceOnUse%22 style=%22mask-type:alpha%22%3E%3Cpath fill=%22%23E40000%22 d=%22M22 2.095C22 .938 21.105 0 20 0H2C.895 0 0 .938 0 2.095V22L11 3.143 22 22V2.095Z%22/%3E%3Cpath fill=%22url%28%23tlg%29%22 d=%22M0 2.095C0 .938.895 0 2 0h9v19.905C11 21.062 10.105 22 9 22H0V2.095Z%22/%3E%3C/mask%3E%3Cg mask=%22url%28%23tlm%29%22%3E%3Cellipse cx=%2211%22 cy=%2211.352%22 fill=%22%23E40000%22 rx=%2210.5%22 ry=%2211%22/%3E%3C/g%3E%3Cdefs%3E%3ClinearGradient id=%22tlg%22 x1=%2211%22 x2=%22-1.938%22 y1=%2222%22 y2=%221.894%22 gradientUnits=%22userSpaceOnUse%22%3E%3Cstop stop-color=%22%23FFCBCB%22 stop-opacity=%22.18%22/%3E%3Cstop offset=%221%22 stop-color=%22%23E40000%22/%3E%3C/linearGradient%3E%3C/defs%3E%3C/svg%3E");background-repeat:no-repeat;background-size:calc(var(--u)*1.1);background-position:left calc(var(--u)*5) bottom calc(var(--u)*1.75)}
.page::before{left:calc(var(--u)*6.6)}
.cover{background-image:none}
.tl{position:absolute;top:calc(var(--u)*3.6);left:calc(var(--u)*5);z-index:2;display:flex;align-items:center;gap:calc(var(--u)*.6);color:var(--fg)}
.tl svg:first-child{height:calc(var(--u)*2.2)}.tl svg:last-child{height:calc(var(--u)*1.5)}
@media print{:root{--pw:16in}html{scroll-snap-type:none}body{display:block}.page{margin:0;break-after:page}}
</style>
</head>
<body style="--book:'{Niche}: Niche Guidelines'">
{CARD: <div class="v"><a href="{video url}"><img src="https://i.ytimg.com/vi/{id}/hqdefault.jpg" alt=""></a><b>{ratio}×</b><a href="{video url}"><span>{title}</span></a><small><a href="{channel url}">{channel}</a> · {views} views · {date}</small></div>
{ratio}: views ÷ the channel's three-month median from step 4, the number that picks winners and flops, on every badge and ratio. Never print TubeLab's averageViewsRatio.
ROW CARD: the same with class="v h", everything after the image link in one <div>, and <p>{why, one line}</p> last when the page asks for it. FLOP: class="x" on the image link crosses it out.
Any other ratio, outside a card: <b class="r">{ratio}×</b>, class="r lo" for a flop's.
PAGE: <section class="page"><p class="k">{Chapter · section}</p><h2>{title}</h2>{body}</section>. Each body below is one page. A page that overflows becomes two pages.}
<div class="chapter" style="--ch:'Cover'">
<section class="page cover">
<a class="tl" href="https://tubelab.net" aria-label="TubeLab"><svg viewBox="0 0 22 22" fill="none"><mask id="tlm" width="22" height="22" x="0" y="0" maskUnits="userSpaceOnUse" style="mask-type:alpha"><path fill="#E40000" d="M22 2.095C22 .938 21.105 0 20 0H2C.895 0 0 .938 0 2.095V22L11 3.143 22 22V2.095Z"/><path fill="url(#tlg)" d="M0 2.095C0 .938.895 0 2 0h9v19.905C11 21.062 10.105 22 9 22H0V2.095Z"/></mask><g mask="url(#tlm)"><ellipse cx="11" cy="11.352" fill="#E40000" rx="10.5" ry="11"/></g><defs><linearGradient id="tlg" x1="11" x2="-1.938" y1="22" y2="1.894" gradientUnits="userSpaceOnUse"><stop stop-color="#FFCBCB" stop-opacity=".18"/><stop offset="1" stop-color="#E40000"/></linearGradient></defs></svg><svg viewBox="0 0 69 15" fill="none"><path fill="currentColor" d="M15.736 9.769V4.131h2.723v9.818h-2.614v-1.783h-.102a2.899 2.899 0 0 1-1.106 1.387c-.512.35-1.136.524-1.873.524-.656 0-1.234-.149-1.732-.447a3.076 3.076 0 0 1-1.17-1.272c-.277-.55-.418-1.208-.422-1.975V4.13h2.723v5.766c.004.58.16 1.037.467 1.374.306.337.718.505 1.233.505.328 0 .635-.074.92-.224.286-.153.516-.379.691-.677.18-.298.266-.667.262-1.106ZM20 13.95V.858h.723a2 2 0 0 1 2 2V5.78h.083a3.6 3.6 0 0 1 .518-.805 2.73 2.73 0 0 1 .895-.69c.37-.188.83-.282 1.38-.282a3.69 3.69 0 0 1 1.982.563c.605.37 1.089.931 1.451 1.681.362.746.543 1.681.543 2.806 0 1.095-.177 2.020-.53 2.774-.35.75-.827 1.32-1.432 1.707-.6.383-1.274.575-2.020.575-.528 0-.978-.087-1.349-.262a2.79 2.79 0 0 1-.901-.658 3.347 3.347 0 0 1-.537-.812h-.121v1.572H20Zm2.666-4.91c0 .584.08 1.093.242 1.528.162.435.397.774.703 1.016.307.239.68.358 1.12.358.442 0 .817-.121 1.124-.364.307-.247.54-.588.697-1.023.162-.439.243-.944.243-1.515 0-.566-.08-1.065-.237-1.495-.157-.43-.39-.767-.697-1.01-.306-.243-.683-.365-1.13-.365-.444 0-.82.118-1.126.352-.302.234-.535.567-.697.997-.162.43-.242.938-.242 1.521ZM35.795 14.141c-1.010 0-1.879-.204-2.608-.613a4.203 4.203 0 0 1-1.674-1.752c-.392-.758-.588-1.655-.588-2.69 0-1.010.196-1.897.588-2.66a4.347 4.347 0 0 1 1.655-1.783c.716-.427 1.556-.64 2.519-.64.648 0 1.25.105 1.809.314a4.013 4.013 0 0 1 1.47.926c.422.414.75.934.984 1.56.235.622.352 1.351.352 2.186a.748.748 0 0 1-.748.748l-5.938.006V8.14l4.123-.09c0-.391-.086-.739-.256-1.041a1.832 1.832 0 0 0-.71-.71 2.024 2.024 0 0 0-1.042-.262c-.413 0-.78.096-1.099.288a2.017 2.017 0 0 0-.741.76c-.18.316-.271.667-.275 1.055v1.604c0 .486.09.906.268 1.26.184.353.441.626.774.818.332.192.726.287 1.182.287.303 0 .58-.042.831-.127.252-.085.467-.213.646-.384.179-.17.315-.38.409-.626l2.518.166a3.438 3.438 0 0 1-.786 1.585c-.392.448-.9.797-1.521 1.049-.618.247-1.332.37-2.142.37ZM42 13.95V.858h.768a2 2 0 0 1 2 2v8.809h2.326c.281 0 .525.193.588.467.197.844.83 1.518 1.66 1.767l.158.047H42ZM51.97 14.135c-.626 0-1.184-.109-1.674-.326a2.696 2.696 0 0 1-1.163-.978c-.281-.435-.422-.976-.422-1.624 0-.545.1-1.003.3-1.374.2-.37.473-.67.819-.895a3.97 3.97 0 0 1 1.176-.511 8.967 8.967 0 0 1 1.393-.243c.571-.06 1.031-.115 1.38-.166.35-.056.604-.137.761-.243a.534.534 0 0 0 .237-.473v-.039c0-.404-.128-.718-.383-.94-.252-.22-.61-.332-1.074-.332-.49 0-.88.109-1.17.326a1.5 1.5 0 0 0-.575.806l-2.519-.205c.128-.596.38-1.112.754-1.547a3.62 3.62 0 0 1 1.451-1.010c.597-.238 1.287-.358 2.071-.358.546 0 1.068.064 1.566.192.503.128.949.326 1.336.595.392.268.701.613.927 1.035.226.418.339.918.339 1.502v6.622h-2.582v-1.361h-.077a2.77 2.77 0 0 1-.633.812 2.93 2.93 0 0 1-.952.543c-.371.128-.8.192-1.285.192Zm.78-1.88c.401 0 .755-.078 1.062-.236a1.88 1.88 0 0 0 .722-.652c.175-.273.262-.582.262-.927V9.398a1.46 1.46 0 0 1-.351.154 6.404 6.404 0 0 1-.493.121c-.183.034-.366.066-.55.096l-.498.07c-.32.047-.599.122-.837.224-.239.102-.424.24-.556.416-.132.17-.199.383-.199.639 0 .37.135.654.403.85.273.192.618.287 1.036.287ZM58.925 13.95V.858h.723a2 2 0 0 1 2 2V5.78h.083a3.6 3.6 0 0 1 .518-.805 2.73 2.73 0 0 1 .895-.69c.37-.188.83-.282 1.38-.282a3.69 3.69 0 0 1 1.982.563c.605.37 1.088.931 1.45 1.681.363.746.544 1.681.544 2.806 0 1.095-.177 2.020-.53 2.774-.35.75-.827 1.32-1.432 1.707-.601.383-1.275.575-2.020.575-.529 0-.978-.087-1.349-.262a2.79 2.79 0 0 1-.901-.658 3.345 3.345 0 0 1-.537-.812h-.122v1.572h-2.684Zm2.665-4.91c0 .584.081 1.093.243 1.528.162.435.396.774.703 1.016.307.239.68.358 1.119.358.443 0 .818-.121 1.125-.364.307-.247.539-.588.697-1.023.161-.439.242-.944.242-1.515 0-.566-.078-1.065-.236-1.495-.158-.43-.39-.767-.697-1.010-.307-.243-.684-.365-1.131-.365-.443 0-.818.118-1.125.352-.303.234-.535.567-.697.997-.162.43-.243.938-.243 1.521Z"/><path fill="currentColor" d="M11.605 1.753a3 3 0 0 1-2.296 1.068H6.596V14.24H4.079V2.82H0V.761h12.44l-.835.992Z"/></svg></a>
<div class="wall">{16 winners: <a href="{video url}"><img src="{hqdefault}" alt=""></a>}</div><div class="shade"></div>
<div><p class="k">{Competitors of <a href="{url}">{channel}</a> · }{run date}</p><h1>{Niche}:<br><span class="mut">Niche Guidelines</span></h1>
<p class="lead">What wins, what flops and how to make it, from {n} winners and {n} flops.</p>
<p class="mut">{n} channels: {n} emerging and {n} established · {n} videos cited</p></div>
</section>
</div>
<div class="chapter" style="--ch:'Research'">
<section class="page divider"><p class="n">01</p><h1>Research</h1><p class="lead mut">The niche · Working formats · The channels · Who watches, and how it pays · Outliers · What to avoid · What got similar channels in trouble · Patterns that work</p></section>
{The niche: <div class="split"><div class="col"><p class="lead">{what it is}</p><dl class="kv"><dt>Who watches</dt><dd>{the most common age, gender, income tier and region across the channel hits, as an estimate}</dd><dt>What wins</dt><dd>{with numbers}</dd><dt>Start with</dt><dd>{the pattern to make first, and why}</dd></dl></div><div class="col">{CARD}</div></div>}
{Working formats: the same split. kv: Format, Length, Faceless or on camera ({n} of {n} winners are on faceless channels, from the hits and checked on the thumbnails), Where to differ. Right: <div class="g2">{four CARDs}</div>}
{The channels, h2 "{n} channels: {n} emerging, {n} established": <div class="chans"><a{ class="me" for the given channel} href="{url}"><img class="av" src="{avatar url}" alt=""><b>{channel}</b>{emerging or established} · {subs} subscribers<br><span class="mut">median {views} views · {ratio} per subscriber</span></a></div><p class="mut"><small>Medians cover each channel's last three months of long videos. Left out: {channels and why}.</small></p>}
{Who watches and how it pays, h2 "Who watches, and how it pays": <div class="split"><div class="col"><p class="k">Who watches</p><dl class="kv"><dt>Age</dt><dd>{most common band} · {n} of {n} channels</dd><dt>Gender</dt><dd>{male, female or balanced} · {n} of {n}</dd><dt>Income</dt><dd>{low, middle, upper-middle or high} · {n} of {n}</dd><dt>Region</dt><dd>{region} · {n} of {n}</dd></dl></div><div class="col"><p class="k">How it pays</p><dl class="kv"><dt>Ads on</dt><dd>{n} of {n} channels</dd><dt>RPM</dt><dd>{band}</dd><dt>Also earn from</dt><dd><div class="chips">{one per detected method (sponsors, affiliate links, products, courses, newsletter, community, services, donations): <div class="chip"><b>{method}</b> · {n} channels</div>}</div></dd></dl></div></div><p class="mut"><small>Estimates from TubeLab's channel search, for {n} of the {n} channels. No estimate for: {channels found only through a video search}. {If the fields come back empty: "not available on this plan".}</small></p>}
{Outliers, two pages of six: <div class="g2 fill">{ROW CARDs with why}</div><p class="mut"><small>{n} of {n} are also outliers against their channel's whole list, by TubeLab's flag. The rest are recent wins only: {titles}.</small></p>}
{What to avoid: <p class="lead">{what the flops share}</p><div class="g2 fill">{six FLOP ROW CARDs with why}</div>}
{What got similar channels in trouble, from the risk searches. Keep only channels that share a sampled title formula or thumbnail style. If none is close, one line saying so: <p class="lead">{n} similar channels were terminated and {n} lost ads in the last year. {what their formats share}</p><div class="g2 fill">{one ROW CARD per channel, up to six, for the video it is cited by: <b class="r lo">{terminated | lost ads {Mon YYYY}}</b> in place of the ratio, a terminated channel's image from https://cdn.tubelab.net/vi/{id}.jpeg, <small> with the channel only, and <p>{its title formula, slots in [ ], and its look}</p> last. Four fit under the lead; the rest and the note go on a page headed "What got similar channels in trouble, continued"}</div><p class="mut"><small>The data shows what happened, never why. AI-made channels found in the niche: {n}, {n} of them classed low effort. They share: {format}.</small></p> Never print this note: a terminated channel's thumbnails load from https://cdn.tubelab.net/vi/{id}.jpeg.}
{One page per pattern, h2 = its name: <p class="mut">{n} winners on {n} channels</p><div class="split"><div class="col"><dl class="kv">{Angle, New and specific, Title (formula, slots in <mark>[ ]</mark>), Thumbnail (composition name), Hook (formula name), Script (video formula name, only when read), Who can copy}</dl></div><div class="col">{three ROW CARDs}</div></div>}
</div>
<div class="chapter" style="--ch:'Packaging'">
<section class="page divider"><p class="n">02</p><h1>Packaging</h1><p class="lead mut">Anatomy of an outlier · Title formulas · Winning words · Thumbnail design</p></section>
{Anatomy of an outlier: <div class="split"><div class="col">{ANNOTATED}</div><div class="col"><p class="f">{its title, the slots in <mark></mark>}</p><p class="mut">{channel · views · date · ratio}</p><dl class="kv">{In the title, In the picture, Always}</dl></div></div>}
{ANNOTATED, boxing what matters on one real thumbnail (face, payout screen, amount, arrow), in percent of the image: <a class="anno" href="{video url}"><img src="{hqdefault}" alt=""><i data-n="1" style="left:{x}%;top:{y}%;width:{w}%;height:{h}%"></i></a><ol class="legend"><li>{what box 1 holds, <strong>words on it</strong>}</li></ol>}
{One page per title formula, no h2: <p class="f">"{formula, slots in <mark>[ ]</mark>}"</p><div class="split"><div class="col"><p><span class="big">{n}</span> winners on {n} channels</p><p><span class="big">{n}</span> flops</p><p class="mut">{Rising across YouTube: {n} videos use it, {ratio}× on average, use up 10% or more · click triggers: {triggers} | Seen in this niche only}</p></div><div class="col">{three ROW CARDs}</div></div>}
{More formulas, four a page: <div class="g2 fill"><div class="col"><p class="lead">"{formula}"</p><p class="mut">{counts}</p>{ROW CARD}</div></div>}
{Worn out, two a page: <div class="col"><p class="lead">"{formula}"</p><p class="mut">{counts}</p><div class="g2">{winner ROW CARD}{FLOP ROW CARD}</div></div>}
{Winning words: <div class="g3 fill"><div class="col" style="grid-column:span 2"><p class="k">Win</p><div class="g2">{CHIPs}</div></div><div class="col"><p class="k">Worn out: as common in flops</p><div class="chips">{CHIPs with class="chip worn"}</div></div></div>
CHIP: <div class="chip"><b>{word or topic}</b> · {n} winners, {n} channels, {n} flops<br><small class="mut">e.g. <a href="{video url}">{title}</a> · <b class="r">{ratio}×</b></small></div>}
{One page per thumbnail composition, h2 = its name: <p class="mut">{n} winners on {n} channels · {n} flops</p><div class="split tight"><div class="col">{ANNOTATED, legend moved right}<small class="mut">{proof}</small></div><div class="col">{legend}<dl class="kv"><dt>Text and colour</dt><dd>..</dd><dt>Why it likely works</dt><dd>..</dd></dl><div class="g2 one">{two CARDs}</div></div></div>}
</div>
<div class="chapter" style="--ch:'Storytelling'">
<section class="page divider"><p class="n">03</p><h1>Storytelling</h1><p class="lead mut">Hook formulas · Video formulas</p></section>
{One page per hook formula, h2 = its name: <p class="mut">{n} of {n} openings read, on {n} channels · the first {n} seconds · Who can copy: {..}</p><div class="beats"><span>{beat}</span><span>{beat}</span></div><div class="g2 fill"><div class="q"><small><a href="{video url}&t=0s">{title}</a> · {channel · views · date · ratio}</small><p><b>{Beat}</b>"{its words}"</p></div></div>}
{One page per video formula, h2 = its name: <p class="mut">{n} of {n} scripts read</p><div class="split"><div class="col"><div class="gantt"><span><strong>{Beat: Hook, Setup, Re-hook, Stakes increase, Peak, Twist or Payoff}</strong> <span class="mut">{what happens}</span></span><div{ class=pay on the main payoff}><i style="--a:{start %};--b:{end %}"></i></div><span></span><span class="mut" style="display:flex;justify-content:space-between;font-size:.8em"><span>0%</span><span>50%</span><span>100% of the runtime</span></span></div></div><div class="col"><dl class="kv">{Main payoff, What repeats, Example: linked title, its stats and length}</dl><div class="chips"><a class="chip" href="{video url}&t={s}s"><b>{m:ss}</b> {beat}</a></div></div></div>}
</div>
<div class="chapter" style="--ch:'Learnings'">
<section class="page divider"><p class="n">04</p><h1>Learnings</h1><p class="lead mut">Do's and don'ts · New angles{ · Gaps to win}</p></section>
{One page per pair, h2 = the lesson: <div class="vs"><div class="v yes"><div><span class="tag">✓ Do</span><b>{ratio}×</b></div><a href="{video url}"><img src="{hqdefault}" alt=""></a><a href="{video url}"><span>{title}</span></a><small>{channel · views · date}</small></div><div class="v no">{the dead version, the same with ✕ Don't and class="x" on its image link}</div></div><p class="lead">{the reason}</p>}
{One page per new angle, two patterns blended, h2 = the angle: <div class="split"><div class="g2"><div class="col"><p class="k">{pattern}</p>{CARD}</div><div class="col"><p class="k">+ {pattern}</p>{CARD}</div></div><div class="col"><dl class="kv">{New angle, Why it should work, Filmed yet? (not in the {n} videos counted)}</dl></div></div>}
{Given a channel: one page per gap, h2 = the gap: <div class="split"><div class="g2">{two competitor CARDs}</div><div class="col"><dl class="kv">{Their proof, Why it fits}</dl></div></div>. Then a page "Overall": kv of Copy, Avoid, Position, each <span class="lead"></span>.}
</div>
<div class="chapter" style="--ch:'Video ideas'">
{Only after video-ideation ran.}
<section class="page divider"><p class="n">05</p><h1>Video ideas</h1><p class="lead mut">{how many ideas, for whom, from what}</p></section>
{Overview: <table class="t"><tr><th>#</th><th>Idea</th><th>From</th><th>Potential</th><th>Fit</th></tr></table>}
{One page per idea, h2 = "{title}". Each takes a title formula, composition and hook from this report: <div class="split"><div class="col"><dl class="kv">{From, Title formula, Thumbnail (composition: its parts for this video), Hook (formula: its beats for this video), Avoids (linked dead version, ratio, why), Potential · Fit}</dl></div><div class="col"><p class="k">Proof</p>{CARD}</div></div>}
<section class="page"><p class="k">Next steps</p><h2>What to run next</h2><div class="col fill"><div><p><code>{skill}</code> <strong>{what to do}</strong></p><p class="mut">"{a prompt to paste, [brackets] for what the user fills in}"</p></div></div><p class="mut"><small>Made with the YouTube research skills on the <a href="https://tubelab.net/youtube-mcp">TubeLab MCP</a> · Install: <code>npx skills add TubeLabHQ/youtube-research-skills</code></small></p></section>
</div>
<script>
const P=[...document.querySelectorAll('.page')];
addEventListener('keydown',e=>{const d={ArrowRight:1,ArrowDown:1,PageDown:1,ArrowLeft:-1,ArrowUp:-1,PageUp:-1}[e.key];if(!d||innerWidth<=760)return;e.preventDefault();const m=scrollY+innerHeight/2,i=P.findIndex(p=>p.offsetTop+p.offsetHeight>m);(P[i+d]||P[i]).scrollIntoView({behavior:'smooth',block:'center'})});
</script>
</body>
</html>- File:
reverse-engineer-niche-{niche-slug or handle}.html, or an artifact. Chat: the top finding and the file. - Plain, short language, with causes stated as likely. One idea and a few lines per page.
- Link every video, channel and moment from the data.
- Spell formulas and compositions out where used, never by page number.
Common Pitfalls
- Do not call one channel's hits a pattern. One viral video proves the idea, not the channel.
- Do not sell a pattern seen only on established channels to a new channel. It may ride on trust.
- Do not say why a channel was terminated or lost ads. The data does not show why.