Hook generation is the part of LinkedIn drafting where the model wants to hedge most, because hooks are short and every word is doing visible work. hook-generator forces specificity by giving the model a tight constraint set and asking for six variations that do *different* jobs rather than six restatements of the same idea. The constraint is what makes the output usable.
Most one-shot hook prompts return six variations of the same hook. hook-generator returns six hooks doing six different jobs. The shape is the same in each case (two lines, forty characters per line, no questions in the opener), but the angle deliberately rotates across the set: number-led, contrarian, personal transformation, authority steal, admission, future shock. The writer picks the angle that fits the post they are about to write, not the line they like the sound of.
§01What it does
The skill takes a topic and emits six numbered hooks, each labelled with its angle. Number-led leads with a specific metric or count. Contrarian states a belief and undercuts it. Personal transformation sets up a before-and-after with a digit attached. Authority steal references a name, brand, or tool the reader recognises. Admission opens with a confessed mistake or loss. Future shock makes a forward-looking claim about what is about to change.
Every hook is two lines, each line forty characters maximum, with line two contradicting or reframing line one. Every variation includes at least one I or How I statement somewhere across the two lines, and a digit where the angle allows. No questions in the opener. No em-dashes. No filler.
§02The angle rotation is the whole skill
The reason this works is simple. A writer asking for "good hooks for this topic" usually has a vague preference for one specific angle and will reject the other five out of habit, but the rejection is information. Seeing six hooks across six rotations forces the writer to commit to which job the hook is doing, not just which line they prefer. That commitment is what turns the hook into a real opener for the post that follows.
The pairing with post-formatter is what most users miss. Generate the six, pick one, then call post-formatter with that hook locked in. The framework already enforces a fifty-character cap on the hook line, so a forty-character hook from this skill drops in cleanly with a six-character buffer for cleanup.
§03Setup
# Trigger phrases:
# "write me hooks"
# "hook ideas"
# "generate hooks"
# "I need a hook for a post about..."
The skill is configured to auto-start with no preamble. The first response after triggering is the six-variant output, not a description of what hooks are.
§04Caveats
Forty characters is short. Some topics need more. The skill will not negotiate the length cap, which means the writer occasionally needs to expand the chosen hook by hand once it is selected. That is a feature: the cap forces the writer to find the highest-density phrasing first, then loosen it deliberately rather than starting loose.