YouTube Viral Optimizer Skill
The YouTube Viral Optimizer skill (skill-viral-videos)
teaches Claude Code, Claude Desktop, OpenClaw, or any other LLM client how to write
YouTube titles, thumbnail concepts, and full video scripts that people actually click on. It encodes concrete CTR rules
(one focal point per thumbnail, tension before explanation, no numbered lists), YouTube algorithm factors
(topic lanes, new-audience signals, timing windows), and a proven script structure that keeps viewers watching.
It's built for channel owners and content teams who want a repeatable method instead of guessing: every title and script is checked against the same rules, and — when the InfraNodus MCP server is connected — against real search demand and content gap data, so your videos target what people search for but can't find yet.
Get Started
Create an Account
Sign up for a free InfraNodus account to get access to the knowledge graph tools and your API key.
Create a Free AccountGet InfraNodus MCP
https://mcp.infranodus.comAdd this URL as a connector via Settings > Connectors in Claude Web / Desktop and authenticate with your InfraNodus API key via OAuth.
https://mcp.infranodus.comOpen the ChatGPT Apps page > Advanced Settings > enable Developer Mode > Create App, paste this URL (leave the OAuth fields empty), click Connect, and authenticate with your InfraNodus API key. Full guide: ChatGPT setup.
claude mcp add infranodus -s user --transport http https://mcp.infranodus.com/mcp
Then open Claude Code in terminal, run /mcp and pick
"Authenticate" — it opens the browser OAuth flow where you will need to
put in your InfraNodus API key to connect to
the MCP server.
{
"mcpServers": {
"infranodus": {
"url": "https://mcp.infranodus.com"
}
}
}
Add this to ~/.cursor/mcp.json (all projects) or
.cursor/mcp.json (one project) — or use
Cursor Settings > MCP > New MCP Server. Then
click "Login" next to the server to authenticate with your
InfraNodus API key.
codex mcp add infranodus --url https://mcp.infranodus.com
codex mcp login infranodus
Run these commands in your terminal — the login command opens the
browser OAuth flow where you enter your
InfraNodus API key. (You can also add the
server to ~/.codex/config.toml manually.)
{
"mcpServers": {
"infranodus": {
"command": "npx",
"args": ["-y", "infranodus-mcp-server"],
"env": {
"INFRANODUS_API_KEY": "your-api-key"
}
}
}
}
Add this to your LLM client's MCP configuration file, replacing
your-api-key with your
InfraNodus API key. See our
deployment guides for client-specific
instructions (n8n, Make.com, local setup, and more).
Install the Skill
/plugin marketplace add infranodus/skills
/plugin install skill-viral-videos@infranodus-skills
Run these two commands in the Claude Code prompt to install
the skill as a plugin (use
infranodus-all@infranodus-skills to install all
our skills at once). Or use npx in your terminal:
npx skills add infranodus/skills --skill skill-viral-videos -g -a claude-code -ynpx skills add infranodus/skills --skill skill-viral-videos -gWhen prompted, select the agents where you want to add the skill — e.g. claude-code, codex, or cursor.
npx comes bundled with Node.js. For Claude Web / Desktop, ChatGPT, and OpenClaw, see the full installation instructions.
Didn't work? You can simply download skill-viral-videos.zip from our GitHub repository and add it to your LLM client manually — see the step-by-step instructions.
What the Skill Does
Once installed, prompts like these trigger the skill's rules and workflows:
- "Give me title and thumbnail options for this video" — generates 3-5 title + thumbnail combinations, each self-rated for tension, concrete language, mobile readability, and algorithm fit.
- "Write a script for this topic" — builds a full script through the fixed stage sequence, with a checklist run before delivery.
- "Why isn't this video performing?" — audits the title, thumbnail, and hook against the CTR rules and the channel's topic lane.
- "Which keyword should I make a video about?" — scores keyword opportunities by demand, competition, structural gap potential, and fit with your channel's differentiator.
- "Optimize this title" — rewrites method-first, abstract, or list-style titles into outcome-first, opinionated ones.
Before generating anything, the skill establishes the channel context: your core topics, audience, how YouTube's algorithm perceives your channel, and your differentiator — the unique lens that should thread through every title and script (e.g. "real-world durability tests, not specs" for a tech channel).
How It Works: The Script Structure
Every script follows the same tension architecture. Two things come first: address a big TAM (open with a universal problem, not niche terminology) and validate credibility (a short trust signal about who you are). Then the stages run in order:
| Stage | Purpose |
|---|---|
| 1. Pattern Interrupt (1-3s) | Stop the scroll — an unexpected statement, strong claim, or visual contrast. |
| 2. Hook (5-7s) | Promise of value: why should the viewer care? |
| 3. Framing (10-15s) | Why this matters to them specifically, plus your credibility line. |
| 4. Curiosity Loop | Open a knowledge gap — reveal that something hidden exists without explaining it yet. |
| 5. Escalation | Raise the stakes — show the problem is bigger than the viewer thought. |
| 6. Payoff | Resolve the loop — deliver the core insight clearly and memorably. |
| 7. Relevance Bridge | "Here's what this means for you" — turn the insight into action. |
| 8. Loop Reopen (optional) | Tease the next question — for retention, series, or looping back to the start. |
Title and thumbnail rules complete the picture: lead with outcome, not method; avoid numbers and abstract words; and make sure the title and thumbnail never say the same thing — together they create a gap the viewer has to click to close.
The Graph-to-Script Pipeline (InfraNodus MCP)
When the InfraNodus MCP server is connected, the skill stops inventing curiosity loops from intuition and extracts them from a knowledge graph of your topic instead. Clusters show what's well-covered; the structural gaps between them are ready-made tension — something the audience senses should be connected but isn't.
| Graph output | Feeds script stage |
|---|---|
Topical clusters (analyze_text) | Framing — what the topic is about |
Structural gaps (generate_content_gaps) | Curiosity Loop — the hidden connection nobody talks about |
Research questions (generate_research_questions) | Hook — the question the viewer needs answered |
Conceptual bridges (develop_conceptual_bridges) | Escalation — how the niche topic connects to universal themes |
Latent topics (develop_latent_topics) | Payoff — the insight viewers can't get anywhere else |
A parallel SEO workflow uses analyze_related_search_queries, analyze_google_search_results, and
search_queries_vs_search_results to find topics with search demand but thin supply — the gap topic becomes the
curiosity loop, the demand language becomes the hook, and your unique answer becomes the payoff. The skill still works without
the MCP server — the CTR and script rules apply either way — but the gap research is where the edge comes from.
Install the Skill
Claude Code
Download the skill into your global skills folder:
cd ~/.claude/skills
curl -L -o skill-viral-videos.zip https://github.com/infranodus/skills/releases/latest/download/skill-viral-videos.zip
unzip skill-viral-videos.zip -d skill-viral-videosThen ask for titles, thumbnails, or a script — Claude will invoke the skill when relevant.
Claude Web / Desktop
Go to Settings > Capabilities, enable Code execution and file creation, scroll to the
Skills section, and upload the skill-viral-videos.zip file.
OpenClaw
Run in the OpenClaw chat:
install this skill: https://github.com/infranodus/skills/releases/latest/download/skill-viral-videos.zipSee the main installation guide for step-by-step instructions per LLM client.