Skills LibraryDevRel
← All skills

hook-engineering

/hook-engineering · View on GitHub

Rewrites the opening hook of a content script using proven, high-retention hook formulas, then explains why each works. Use when you have a video, short, or social script and want a stronger first line that stops the scroll without overpromising.

devrelcontentvideocopywritinghooks

Run this in your terminal — it installs into ~/.claude/skills:

curl -fsSL https://raw.githubusercontent.com/SamRome1/devrel-skills/main/install.sh | bash -s -- hook-engineering

Pass --projectbefore the skill name to install into the current project's .claude/skills instead.

Hook Engineering

You take an existing content script and engineer a better hook — the first 1–2 lines (roughly the first 3 seconds of a video). The hook's only job is to buy the next sentence. A great script with a weak hook dies in the feed.

You do not rewrite the whole script. You diagnose the current hook, then produce a few stronger alternatives built from proven formulas, each labeled so the author understands the move and can reuse it.

When to use this

  • The user pastes a script (YouTube, Short/Reel/TikTok, LinkedIn, X thread) and asks for a better opener, a stronger hook, or "make this stop the scroll."
  • A draft opens with throat-clearing ("Hi everyone, in today's video…"), a slow setup, or a vague promise.

Inputs to collect (ask only if missing)

  1. The script (or at least the intended topic + payoff).
  2. Platform — short-form video, long-form video, or text post. This changes length and pacing. If unstated, assume short-form video.
  3. The payoff — what the viewer actually gets by the end. The hook must promise this and nothing more.

Process

  1. Isolate the current hook. Quote the first 1–2 lines verbatim so the author sees exactly what you're judging.
  2. Diagnose it against the quality bar below. Name the specific failure (e.g. "buried the payoff behind a greeting," "abstract — no concrete stakes").
  3. Pick 2–3 formulas from references/hook-formulas.md that fit this content. Don't force a formula that the script can't deliver.
  4. Write 3–5 alternative hooks. For each: the rewritten line(s), the formula name, and one sentence on why it works here.
  5. Flag the promise. Confirm each hook is backed by the script's actual payoff. If a hook overpromises, say so and either soften it or note what the script would need to add to earn it.
  6. Recommend one as your top pick, and say why.

The quality bar

A strong hook is:

  • Specific, not generic. "I cut our query time by 90%" beats "let's talk about performance." Numbers, names, and concrete nouns earn attention.
  • Tension-loaded. It opens a curiosity gap, names a pain, or makes a claim the viewer needs resolved.
  • Front-loaded. The most interesting word lands in the first few words. No greetings, no "in this video," no slow runway.
  • Audience-aimed. The viewer should feel "this is about me / my problem."
  • Honest. It promises only what the script pays off. Clickbait that doesn't deliver trains the audience to scroll past you next time.
  • Short. One breath. If you can't say it before the viewer's thumb moves, it's too long.

Output format

Current hook:
> "<verbatim opening>"

Diagnosis: <what's weak, named specifically>

Rewrites:
1. "<hook>"  — [Formula] — <why it works here>
2. "<hook>"  — [Formula] — <why it works here>
3. "<hook>"  — [Formula] — <why it works here>

Top pick: #<n> — <one-line rationale>
Promise check: <does the script deliver on these? any overpromise to fix?>

Gotchas

  • Match the payoff, not the dream. The fastest way to lose trust is a hook the content doesn't earn.
  • Keep the author's voice. Engineer the structure; don't flatten their tone into generic hype.
  • Platform matters. A LinkedIn first line and a TikTok first second are not the same length or register — see the per-platform notes in the formulas file.
  • One idea per hook. If the hook needs an "and," it's two hooks. Pick one.