Index of every page on this site: llms.txt
All skills
sergebulaev avatar

/yt-hook-scripter

@f422cea
Run it yourself

Script the spoken opening that decides retention: the first 30 seconds of a long-form YouTube video or the first 3 seconds of a YouTube Short. Uses 2026 patterns (restate-and-raise, cold-open payoff tease, question-and-contract, and the 3-second frame-one Shorts hook with a designed loop) to confirm the title's promise, open a loop, and earn the next 30 seconds with no intro. Use to write or fix a video opening. Not for the title (use yt-title-optimizer) or description (use yt-description-writer).

Use this Skill: https://skilld.dev/gh/sergebulaev/youtube-skills/yt-hook-scripter

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈130 tokens for metadata: the name and description. ≈1.5k when used: this file. ≈836 more on demand in 1 file.

Description uses 6.5% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈521 characters, or 6.5%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

YouTube Hook Scripter

The packaging earns the click; the first 30 seconds (or 3 on a Short) earns the watch-time. This skill scripts that opening so the retention graph starts as a slope, not a cliff. It writes the words you say on camera, not the metadata.

When to use

  • User says "write my intro" or "how should this video open"
  • A video's retention drops hard in the first 30 seconds and needs a re-cut
  • User has a Short and needs the first 3 seconds to stop the swipe
  • User wants the spoken hook to match the title's promise

Formulas this skill uses (opening-hook shapes)

Code Formula Surface Best for
Y7 Restate-and-Raise long-form 30s confirm the promise, raise the stakes
Y8 Cold-Open Payoff Tease long-form 30s flash the result, cut to the setup
Y9 Question-and-Contract long-form 30s answer fast, reopen a deeper loop
Y10 Shorts 3-Second Hook Shorts 3s state payoff/tension on frame one, loop the end

Full skeletons in ../../references/hook-formulas.md.

Steps

Voice profile first (all drafts). If ../../references/voice-profile.md has filled: yes, load it and match the user's voice fingerprint, hard rules, and CTA/link style throughout. If it is not filled, mention once that yt-community-post-writer --mode profile can learn their voice from a few posts, then proceed with the generic voice rules.

Never invent the specifics. The rules below ask for a concrete number, a date and a named entity, because that is what separates a real post from a generated one. Take them from what the user actually said in this conversation. Do not invent a figure, a date, a client name or a result, and do not soften a vague claim into a plausible-looking number. If the user has nothing concrete for a beat, ask them once, and if they still have nothing, drop the claim rather than decorate it. A published invented number is a retraction; a missing one is only a weaker post.

  1. Gather inputs. The title (the promise to keep), the single payoff of the video, the most dramatic moment or result available for a cold open, the target viewer, and whether it is long-form or a Short. If a URL is given, parse it with lib/url_parser.py; is_short decides the surface.
  2. Pick the surface.
    • Long-form: write the first 30 seconds, beat by beat, with rough timestamps.
    • Short: write the first 3 seconds (frame one), then the line that makes the ending loop back to the start.
  3. Pick the formula. Use the pairing table in the hook-formulas reference: the title formula suggests the opening hook (a curiosity-gap title pairs with restate-and-raise; a how-I title pairs with a cold-open payoff tease).
  4. Write the script. Spoken voice, one person to one person. For long-form:
    • Line 1 lands the title's promise in your own words (no greeting).
    • Line 2 raises the stakes or names the obstacle (opens the loop).
    • Line 3 makes the watch-contract ("by the end you will..").
    • Note where B-roll, a cold-open clip, or on-screen text goes. For a Short: one arresting frame-one line, on-screen text included, plus the loop line.
  5. Strip the intro. Cut any "welcome back", logo sting, or channel-trailer reflex. The hook is the first thing the viewer hears.
  6. Humanizer pass. Scrub 2026 AI vocabulary by density (one marker in a beat is a slip; three is a script, rewrite the beat), break stacked or hollow rule-of-three, strip reveal bridges ("Here's what nobody tells you") and sincerity openers ("not gonna lie", "let me be honest"); use .. for a spoken pause instead of a dash on the teleprompter. Make it sayable; never insert a punch line or a "No X. No Y. Just Z." beat for rhythm, and never add a hedge the creator did not write.
  7. Approval card. Show the scripted open with timestamps (long-form) or the 3-second frame plan (Short), and the title it is keeping.
  8. On approval. This is a script the user performs; it is not published by itself. If the user uploads through this bundle, the hook is spoken in the video and the title/description ship via lib.publish(...).

Hard rules

Global voice rules: see root SKILL.md Voice rules. Additional skill-specific rules:

  • No greeting, no intro animation, no logo sting before the hook. Ever.
  • Long-form: the promise from the title is restated in the first 1 to 2 spoken lines, then a loop is opened. Earn the next 30 seconds by the 30-second mark.
  • Short: the payoff or tension lands on frame one (second 0 to 1), with on-screen text so it works muted. Design the ending to loop back to the start.
  • The opening can never promise more than the video delivers. Honor the click.
  • Use .. for a spoken pause. A dash in a script is only a breath mark the reader sees, not a tell, but .. reads better on a teleprompter.

Anti-patterns (skill will refuse)

  • "Hey guys, welcome back to the channel" or any greeting in the open.
  • A 15-second intro animation before the first word of value.
  • A cold open that teases a moment not actually in the video.
  • "Make sure to like and subscribe" inside the first 30 seconds.
  • A "No X. No Y. Just Z." staccato stack or "The result?" reveal inserted for punch.
  • Announced candor ("let me be honest", "not gonna lie") with no specific fact behind it.
  • A Short that opens with a slow zoom or a logo instead of the payoff.

Resources

  • ../../references/hook-formulas.md - Y7-Y10 opening-hook skeletons and the title pairing
  • ../../references/algorithm-heuristics.md - the first-30-seconds and 3-second retention rules
  • references/retention-beats.md - the beat-by-beat opening structure for long-form and Shorts

Related skills

  • yt-title-optimizer - the promise the hook keeps
  • yt-thumbnail-brief - the emotion the open should match
  • yt-description-writer - the written hook (different surface from the spoken one)

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts1mo3 checks · Risk SAFE
  • Gen Agent Trust Hub1mo

    The skill scripts YouTube video openings by parsing external URLs and utilizing internal tools for publishing. While these features are core to its purpose, processing external content introduces a minor surface for indirect prompt injection, partially mitigated by a human approval step before publishing.

  • Socket1mo

    No alerts

  • Snyk1mo

    Risk: LOW · No issues

Signed by skilld at f422cea. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 2 months ago

README badge

README badge for sergebulaev/youtube-skills/yt-hook-scripter