YouTube Titles in 5 Minutes: 10-Hook SEO Booster

It's late, the edit is finally done, and you're staring at the title field. This turns that 30-minute "what do I even call this" slump into a structured 5-minute job.

• Type one line about your video → get 10 title candidates sorted by hook type + description + tags + thumbnail text + pinned comment + an A/B testing plan, all in one pass
• Already published? Paste the URL → it pulls the real metadata, runs a 7-point diagnostic checklist, and attaches a rewritten version to every item marked "needs work"
• Judgment, not vibes → the full documentation of 10 title hook types (definitions, examples, writing rules) is included, so you get better at titles the more you use it

📦 Real content preview

This is a new listing with no reviews yet, so here are verbatim excerpts from the included documents (translated from Korean — see the FAQ on language). This density holds across the entire package.

① The "Loss Aversion" entry from the hook-type doc (references/title-hooks.md)

4. Loss Aversion

Make concrete what the viewer loses by not watching. Loss is a stronger click motivator than gain.

• Example: "The side-dish habit costing solo renters an extra ₩100,000 every month"
• Example: "Skip this setting and your battery lifespan takes the hit"
• Rule: the loss must be backed by real evidence; fear-mongering exaggeration is banned.

All 10 types follow this format: definition, 2 examples, 1 writing rule. It's not a doc of examples to copy — it's a set of criteria to apply to your own topic.

② The verdict criteria for title A/B tests (references/ab-test-guide.md)

4. Verdict: Compare CTR only between periods with similar impression volume. If CTR goes up but average view duration drops sharply, that's the signature of a clickbait title — roll it back.

Not "CTR went up, ship it" — the guide tells you when to roll back.

③ Cautions for interpreting results (references/ab-test-guide.md)

• Never compare CTR directly across different impression volumes. As impressions grow, the share of non-subscribers and browse-feed traffic grows, so CTR naturally falls. CTR comparisons between periods with a 2x+ gap in impressions are treated as invalid.

This is exactly the trap most creators fall into ("changed the title, CTR dropped, must be a failure") — and the guide blocks it.

📂 What's included (4 files)

One generation request returns: 10 title candidates (at least 5 distinct hook types, each labeled with its type plus a one-line rationale) · a description with keywords placed in the first 125 characters (chapter timestamps + CTA included) · 10–15 tags (broad + long-tail mix, copy-ready format) · 3 thumbnail text options (labeled by tone: urgency / curiosity / gain) · a pinned comment draft · an A/B testing plan tailored to this video.

The details are built to working standards too. For Shorts, the Shorts ruleset kicks in automatically (title within 40 characters, max 3 hashtags), and thumbnail text is designed to complement the title rather than repeat it (title = what, thumbnail = why now). If yt-dlp is missing or the network is down, only diagnosis and competitor analysis are skipped — topic/script-based generation still works, and the report states what was skipped.

🎬 Three usage scenes

Scene 1 — New video, right before upload

"Next video topic: '5 side dishes that last a week on ₩10,000, for people living alone.' The script is in banchan-script.md. Give me the title, description, and thumbnail text — everything."

→ Returns 10 titles across hook types, a description with chapter timestamps extracted from the script's sections, a tag set, 3 thumbnail text options, a pinned comment, and an A/B plan that says "start with Test & Compare on the 3 thumbnail options."

Scene 2 — Diagnosing an underperforming video

"This video is getting fewer views than I expected. Diagnose whether the title or description is the problem, and compare it against the title patterns of the top 'Galaxy S26 review' videos."

→ Pulls the actual title, description, and tags via yt-dlp, builds a pass/needs-work table across 7 items (title length, hook presence, keywords in the first 125 characters, chapters, tag composition, and more), and attaches a rewrite for every failing item. Then it analyzes the top 15 competitor titles (average length, share using numbers, emotional vocabulary) and proposes improvements in two directions: pattern-conforming and pattern-breaking.

Scene 3 — Cleaning up the back catalog

"Take these 5 videos from my channel and diagnose the titles and descriptions in one go. Which should I fix first?"

→ Batch-fetches all URLs, builds a per-video diagnostic table, and proposes a fix order using the prioritization logic from ab-test-guide.md (start with videos whose CTR is low relative to impressions).

👍 A great fit if you are

• A solo creator who burns 30+ minutes on titles after every edit, right when you're most tired
• A new channel owner with no baseline for how descriptions and tags should even be written
• A brand-channel marketer who needs consistent metadata quality across many videos
• An education/tutorial creator whose search traffic makes chapter timestamps essential
• Already on Claude Code and want your YouTube workflow to stay in the terminal

👎 Not for you if

• You want guaranteed views or impressions — this is a metadata quality tool, and it does not promise algorithmic outcomes. The skill's own hard rules ban performance-guarantee language.
• You expect automatic uploads or edits to YouTube Studio — applying changes is always done by you.
• You expect thumbnail image files — it writes thumbnail text; it does not create images.
• You need live search-volume numbers or internal YouTube Analytics data — it does not access them.

❓ Trust FAQ

Q1. What do I get, and how do I use it?
Right after payment you download the skill package (SKILL.md + 2 reference docs + 1 Python script). Drop the folder into your Claude Code skills directory and you're done — from then on, requests like "give me titles for this video" make Claude automatically follow this skill's workflow and hook-type docs. No installer, no account signup.

Q2. Do I need AI (Claude Code)? And what about language?
The automated workflow runs in Claude Code. The lookup tool itself is language-agnostic: fetch_meta.py works with YouTube videos in any language, worldwide. The reference documents (hook types, A/B guide) and the script's console messages, however, are written in Korean. Source content is written in Korean. When you use it through Claude (or any AI), ask for it in your language — the skill instructs the AI to serve and adapt content in your conversation language. Only the existing-video diagnosis and competitor analysis need the free open-source yt-dlp (brew install yt-dlp); topic/script-based generation needs no external tools, API keys, or logins.

Q3. How do updates work?
When hook types are added, the checklist is refined, or the script improves, the same product page is updated — buyers re-download the latest version at no extra charge. The same applies if YouTube's search layout or Studio features change and the documented criteria need revising.

Q4. Can I get a refund?
Refunds for digital downloads follow the Capafy platform's refund policy. Before buying, please judge the quality directly from the "Real content preview" excerpts above — the density you see there is the density of the whole package.

Q5. Can I use it commercially, for a company channel or client work?
Yes. Everything the skill generates — titles, descriptions, tags — can be used without restriction for personal channels, company channels, and client deliverables. Only reselling or redistributing the skill files themselves is prohibited.

💳 Purchase info

$3 · one-time payment · unlimited use

If titles cost you 30 minutes per video, this is the price of a coffee to make that 5 minutes, every upload. Pay → download → copy into your skills folder — three steps, and it's live for your next upload.

This product is a metadata quality tool and does not guarantee any increase in views or impressions.