Skip to main content
skill CX Codex

audit-channel-risk — Claude Code skill for YouTube channel audits

The audit-channel-risk Claude Code skill scores a YouTube channel’s risks of demonetization and termination. It checks channel content against YouTube policies and looks for similar channels.

MIT ★ 288 updated 2026-11-15

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add TubeLabHQ/youtube-research-skills
/plugin install youtube-research@tubelab
Open on GitHub

What audit-channel-risk does

The skill assesses whether a YouTube channel risks losing monetization or being terminated. It rereads YouTube’s monetization policies, then checks popular and recent videos, the About text, titles, descriptions, thumbnails, and transcripts. TubeLab MCP and an accessible policy page are required; if the page cannot be opened, the skill asks for its text.

It gives separate scores from 0 to 100 for demonetization and termination risk. The assessment uses policy findings and similar channels, including termination and monetization status. The result is an HTML report with scores and supporting evidence. For multiple channels, it also compares shared templates, presenters, and description text.

What it helps with

  • assess a channel’s risk of losing YouTube monetization
  • identify signals linked to channel termination
  • check videos against YouTube monetization policies
  • find similar channels that were terminated or demonetized
  • create an HTML risk report with scores and evidence

Example prompts for the agent

  • Assess my YouTube channel’s demonetization and termination risks.
  • Check the channel’s latest and most popular videos against YouTube monetization policies.
  • Compare these YouTube channels and create an HTML risk report.

How to install audit-channel-risk

1Plugin marketplace

/plugin marketplace add TubeLabHQ/youtube-research-skills
/plugin install youtube-research@tubelab

This installs the whole youtube-research plugin with all its skills

2Skills CLI (Vercel)

npx skills add TubeLabHQ/youtube-research-skills --skill audit-channel-risk -a claude-code

3Manually: put the files in place

~/.claude/skills/audit-channel-risk/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/audit-channel-risk/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about audit-channel-risk

How do I install audit-channel-risk in Claude Code?

Run /plugin marketplace add TubeLabHQ/youtube-research-skills, then run /plugin install youtube-research@tubelab.

What does audit-channel-risk do?

It scores a YouTube channel’s demonetization and termination risks separately. The skill checks channel content and similar channels, then produces an HTML report with scores and supporting evidence.

What does audit-channel-risk need to run?

It requires TubeLab MCP and access to YouTube’s monetization policy page. If the page cannot be opened, the skill asks for its text.

Source text

SKILL.md · MIT
---
name: audit-channel-risk
description: Use when the user wants to know whether a YouTube channel is at risk of being demonetized or terminated, with a score for each, the proof and what to fix first. Requires the TubeLab MCP.
license: MIT
metadata:
  author: tubelab
  version: "1.0"
  requires: tubelab-mcp
---

# Audit Channel Risk

## Overview

YouTube's [channel monetization policies](https://support.google.com/youtube/answer/1311392?hl=en) decide which channels keep monetization, and they change often. Read the page on every run and check the channel against it, never against memory.

## Workflow

1. Read the policy page and list the sections that set a rule for what videos contain. Skip those about the review, the creator's behaviour or another page. If it will not open, ask for a paste.
2. Read each channel for free: its top and newest videos and About text.
3. Check the channel against each section, across titles, descriptions, thumbnails and three to five transcripts. View the thumbnails of the videos you cite. For the section on repetitive content, a video counts when:
   - its script could swap names and places with an earlier video's and still work, or repeats its entries. A shared intro or outro does not count.
   - its title has the shape used on over half the videos. Count shapes, not words.
4. Note what already shows real people behind the channel: a real presenter on camera, and an About text naming who makes the videos.
5. Run both searches below per channel, seeded with it, without the quality filter. Lost AdSense means `monetization.adsense` is false with a `monetizationChangedAt` date. False alone only means not monetized.
6. Rate every hit but the channel itself from 0 to 100: titles first, then thumbnails for title matches, then scripts if still unsure. A close one, 60 or more, shares the title formula and subject, and the thumbnail style. A terminated channel's thumbnails load from `https://cdn.tubelab.net/vi/<id>.jpeg`.
7. Score two risks from 0 to 100, where higher is riskier. Take the highest signal, never a sum. Under 30 is low 🟢, 30 to 59 is medium 🟡, and 60 or more is high 🔴.
   - Termination: 15 points per close similar channel terminated, up to 100.
   - Demonetization: the higher of two signals.
     - Similar channels: 15 points per close one terminated or demonetized, up to 100.
     - Guidelines: its worst section. A section scores 0 when at most two videos, or under a tenth, run into it, 60 when more do, and 100 when over half do. Drop it one level, 100 to 60 or 60 to 30, when the newest ten videos run into it at under half the older ones' rate.
8. With several channels, compare them. A shared template, presenter or description block, or uploads minutes apart, link channels as one network.

Terminated, then lost AdSense:

```json
{ "relatedChannelId": ["<channelId>"], "terminated": true, "publishedAtFrom": "year", "size": 40 }
{ "relatedChannelId": ["<channelId>"], "monetizationChanged": true, "monetizationAdsense": false, "publishedAtFrom": "year", "size": 40 }
```

## Output Format

One HTML page, from this template. Change nothing in the style block.

```html
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<title>Channel Risk Audit: {channel}</title>
<style>
:root{--bg:#fff;--fg:#1c1c1c;--mut:#666;--line:#e3e3e3;--low:#0c8a4d;--med:#c26a00;--high:#d42a2a}
@media(prefers-color-scheme:dark){:root{--bg:#0f0f0f;--fg:#f1f1f1;--mut:#9a9a9a;--line:#2b2b2b;--low:#3ddc97;--med:#ffa400;--high:#ff4e42}}
*{box-sizing:border-box}
body{margin:0;background:var(--bg);color:var(--fg);font:16px/1.5 system-ui,sans-serif}
main{max-width:920px;margin:auto;padding:32px 16px 72px}
a{color:inherit}
h1{font-size:24px;margin:0}
h2{font-size:19px;margin:44px 0 6px}
h3{font-size:13px;margin:16px 0 0;color:var(--mut);text-transform:uppercase;letter-spacing:.06em}
.mut{color:var(--mut)}
.center{text-align:center}
.low{--c:var(--low)}.med{--c:var(--med)}.high{--c:var(--high)}
.gauges{display:flex;justify-content:center;gap:40px;flex-wrap:wrap;margin:40px 0 24px;text-align:center}
.gauge{position:relative;width:200px;font-weight:600}
.gauge svg{display:block;width:132px;margin:0 auto 8px;transform:rotate(-90deg)}
.gauge circle{fill:none;stroke:var(--line);stroke-width:8}
.gauge .arc{stroke:var(--c);fill:color-mix(in srgb,var(--c) 12%,transparent)}
.gauge b{position:absolute;top:0;left:0;width:100%;line-height:132px;font-size:44px;color:var(--c)}
.legend{display:flex;justify-content:center;gap:20px;flex-wrap:wrap;font-size:13px}
.legend i,.row i{flex:none;display:inline-block;width:10px;height:10px;border-radius:50%;background:var(--c)}
.big{font-size:19px;max-width:660px;margin:24px auto 8px}
.cols{display:grid;grid-template-columns:repeat(auto-fit,minmax(300px,1fr));gap:0 44px}
.row{border-bottom:1px solid var(--line)}
div.row,.row>summary{display:flex;align-items:center;gap:12px;padding:13px 0}
.row>summary{cursor:pointer;list-style:none}
.row>summary::-webkit-details-marker{display:none}
.row>summary::after{content:"+";color:var(--mut);width:12px;text-align:center}
.row[open]>summary::after{content:"\2212"}
.row span{flex:1}
.row b{color:var(--c)}
.row small{display:block;font-size:13px;color:var(--mut)}
.proof{padding:0 24px 18px 22px}
.proof p{margin:6px 0}
.vids{display:flex;gap:14px;flex-wrap:wrap;margin:10px 0}
.vids a{width:200px;font-size:13px;line-height:1.35;text-decoration:none}
.vids img{display:block;width:100%;aspect-ratio:16/9;object-fit:cover;border-radius:8px;margin-bottom:6px;background:var(--line)}
blockquote{margin:10px 0;padding-left:12px;border-left:3px solid var(--line);color:var(--mut)}
ol{padding-left:22px}
li{margin:12px 0}
table{border-collapse:collapse;width:100%;margin-top:24px}
td,th{text-align:left;padding:10px 12px 10px 0;border-bottom:1px solid var(--line)}
</style>
</head>
<body>
<main>
{With several channels, one page: first a table of channel, both scores and biggest risk, then one line on what links the channels (a shared template, presenter, description block or upload times), then everything below once per channel, riskiest first.}

<h1>Channel Risk Audit: <a href="{url}">{channel}</a></h1>
<p class="mut">{run date} · checked against <a href="{policy page url}">YouTube's channel monetization policies</a></p>

<div class="gauges">
  <div class="gauge {low, med or high}"><svg viewBox="0 0 120 120"><circle cx="60" cy="60" r="54"/><circle class="arc" cx="60" cy="60" r="54" pathLength="100" stroke-dasharray="{n} 100"/></svg><b>{n}</b>Demonetization risk</div>
  <div class="gauge {band}"><svg viewBox="0 0 120 120"><circle cx="60" cy="60" r="54"/><circle class="arc" cx="60" cy="60" r="54" pathLength="100" stroke-dasharray="{n} 100"/></svg><b>{n}</b>Termination risk</div>
</div>
<div class="legend"><span class="low"><i></i> 0–29 low</span><span class="med"><i></i> 30–59 medium</span><span class="high"><i></i> 60–100 high</span></div>
<p class="big center">{biggest risk, one sentence with numbers}</p>
<p class="mut center">Higher is riskier. The highest signal sets each score. A low score means no signal was found, not that the channel is safe.</p>

<h2>Signals</h2>
<div class="cols">
  <div>
    <h3>Demonetization</h3>
    <div class="row {band}"><i></i><span>Similar channels terminated or demonetized<small>{n} close of {n} checked</small></span><b>{n}</b></div>
    <div class="row {band}"><i></i><span>Guidelines<small>{n} of {n} sections failed {a section fails when it scores above 0}</small></span><b>{n}</b></div>
  </div>
  <div>
    <h3>Termination</h3>
    <div class="row {band}"><i></i><span>Similar channels terminated<small>{n} of the {n} close ones</small></span><b>{n}</b></div>
  </div>
</div>

<h2>Guidelines check</h2>
{One block per section checked, failed ones first.}
<details class="row {band}">
  <summary><i></i><span>{section name as the page words it}</span><b>{n}</b></summary>
  <div class="proof">
    <p>{what was found, one or two sentences, then how many of the newest ten videos still do it against the older ones, and whether that lowered the score}</p>
    {The next three only for a failed section.}
    <div class="vids">
      <a href="{video url}"><img src="{thumbnail: the 320px mqdefault.jpg, or the TubeLab one for a terminated channel}" alt="">{title} · {views, ratio, date}</a> {up to three videos}
    </div>
    <blockquote>{the policy line it runs into, quoted from the page}</blockquote>
    <p><strong>Fix:</strong> {one sentence per thing to change}</p>
  </div>
</details>

<h2>What happened to similar channels</h2>
{One block per close one, closest first, up to ten. The band is high for terminated, med for lost AdSense. If none is close, one line saying so and naming the nearest.}
<details class="row {band}">
  <summary><i></i><span><a href="{url}">{channel}</a> · {terminated, last upload {date} | lost AdSense {date}}</span><span class="mut" style="flex:none">closeness {n}</span></summary>
  <div class="proof">
    <p>{what matched: the title formula, the thumbnail style or the script structure}</p>
    <div class="vids">
      <a href="{video url}"><img src="{thumbnail}" alt="">{title}</a> {one or two of its videos that match}
    </div>
  </div>
</details>
<p class="mut">{Only if a close one is not counted: which one and why, in one line.}</p>

<h2>Actions to reduce risk</h2>
{One action per fix in a failed section first, biggest first. Then one action for what step 4 found missing: a real presenter on camera, in a video presenting the channel and how its videos are made, an About text naming who makes them, or both. Never an action the channel already does. Word each action as one plain thing to try, clear without its reason, like "Try new title formulas" or "Add a real presenter on camera". Up to four.}
<ol>
  <li><strong>{Action}</strong><br><span class="mut">{why, one sentence}</span></li>
</ol>
</main>
</body>
</html>
```

- File: save it as `channel-risk-{handle}.html` and open it, or make it an artifact.
- Chat: reply with only the two scores, the biggest risk and the file.
- Language: plain words and short sentences. Say "similar channels" and "terminated".
- Proof: every claim has its videos or channels under it.
- Links: link every video and channel you cite, from IDs in the data.

## Common Pitfalls

- Do not say why a similar channel was terminated. The data does not show why.
- Do not rate closeness by topic alone.
- Do not count a later clone of the channel's own videos. Other channels in the same format count.
- Do not flag AI use or upload pace alone.
- Do not score the channel's own AdSense status. Off can mean it was never monetized.
- A score is not a probability, and a low one does not mean safe.
- Never suggest hiding from detection. Every action changes the videos or shows who makes them.

© tubelab · MIT · TubeLabHQ/youtube-research-skills

Similar skills

SKILL.md skills catalog →
partner-channel-program Maya-Beth Finotti partner-channel-program designs the architecture of a B2B or B2C partner program from scratch, testing readiness before picking motions, economics, and governance. CX Codex +1 17,200installs partner-channel-conflict Maya-Beth Finotti partner-channel-conflict drafts rules of engagement for contested deals between direct sales and partners. Covers deal registration, compensation neutrality, and adjudication. CX Codex +1 17,200installs tech-podcast-youtube-channel Samuel Berthe tech-podcast-youtube-channel helps event organizers run a permanent podcast or YouTube channel across editions. Defines off-season ownership, guest vetting pipelines, and handovers. CX Codex +1 6,800installs acquisition-channel-advisor deanpeters acquisition-channel-advisor evaluates acquisition channels through unit economics, customer quality and scale potential. Use it when deciding whether to grow, test or stop a channel. CX Codex +1 2,800installs service-digital-engagement-channel-configure forcedotcom The service-digital-engagement-channel-configure Claude Code skill builds and deploys Salesforce Messaging Channels for In-App and Web. It handles MIAW setup, routing, channel options, and activation. CX Codex +1 2,500installs setup-slack-channel CopilotKit setup-slack-channel — a skill to connect a local CopilotKit Channels agent to Slack from scratch: creates a Slack app, generates tokens, and binds them to a managed Intelligence Channel. Use it to make your bot respond to Slack mentions. CX Codex +1 1,600installs

From the same set23

The whole set →
sponsorship-prospecting The sponsorship-prospecting Claude Code skill finds YouTube channels for product sponsorships and ranks them by audience fit. It activates for creator research and prospect-list building. CX Codex +1 ★ 288repo stars thumbnail-concepts This Claude Code skill develops YouTube thumbnail concepts and studies visual patterns in a niche. It runs on requests for thumbnail ideas or niche research. CX Codex +1 ★ 288repo stars niche-finder The niche-finder Claude Code skill finds YouTube channel concepts and ranks them by opportunity. It activates for requests about open niches or fresh channel formats. CX Codex +1 ★ 288repo stars niche-bending The niche-bending Claude Code skill finds YouTube formats that can move into a different market or language. It starts from a channel the user names or from channels that already earn. CX Codex +1 ★ 288repo stars niche-validator The niche-validator Claude Code skill evaluates whether a YouTube niche is worth entering. It compares market size, saturation, and revenue potential across sampled channels. CX Codex +1 ★ 288repo stars reverse-engineer-channel The reverse-engineer-channel Claude Code skill studies a YouTube channel to find patterns behind its videos. It examines formats, titles, thumbnails, scripts, and revenue sources. CX Codex +1 ★ 288repo stars tubelab-mcp This Claude Code skill explains TubeLab tools, filters, credits, and REST API behavior. It activates when a call fails, returns no results, or someone asks how TubeLab works. CX Codex +1 ★ 288repo stars youtube-research The youtube-research Claude Code plugin researches YouTube niches, channels, trends, and videos. Six agency roles produce reports and ideas for content and monetization. ★ 288repo stars

Project sponsors

Companies that keep this analytics open