View the Social Hook Optimizer SKILL.md the file this guide runs on
---
name: social-hook-optimizer
version: 1.1.0
description: >
Use when the user says "write hooks for this," "fix this hook," "will this
stop the scroll," "optimize this post," "give me openers for this," or
"social copy for [thing]." Rewrites the first line of a social post, scored
against your own post history rather than generic advice. Ships with the
hook rules built in — nothing else to install.
---
# Social Hook Optimizer v1
Rewrites the one line a reader sees before "see more."
## Sibling Skill, Different Baseline
This is the social counterpart to `email-open-optimizer`. Same method, different evidence.
**The two are not interchangeable.** A subject line and a hook are scored against different baselines because the surfaces behave differently. The clearest proof is in the data — the Pamela Oberg announcement was a strong LinkedIn post and the worst-performing SLAI Effect subject line in the same week. Never port a line from one to the other without re-scoring it.
## The Feed Exemption
The subject-line carve-out that `email-open-optimizer` grants does **not** extend here. A social post is public body copy, so whatever banned-word or brand list your organization keeps applies in full to the hook and to everything under it. If you have one, read it fresh and enforce it. If you have none, the Hook Rules below still apply.
The one thing this skill borrows from its sibling is the method, not the exemption.
## Which Skill, When
- Writing a **LinkedIn / Instagram / Facebook / Threads post** — run your voice skill first if you have one, then this skill on the first line.
- Writing an **email subject line, preview text, or first line** — use `email-open-optimizer` instead. Different surface, different baseline, never share lines between them.
## Context Required
This skill runs on its own. Everything it needs is in this folder.
- `linkedin-baseline-30.md` (sibling file) — the scoring baseline. **Required.** It ships with our 30 as a worked example and a header on how to replace them with yours. Score against your own posts, not ours.
- The Hook Rules below — built in, always on.
Read if you have them, skip if you don't.
- Your voice reference (at She Leads AI that is `slai-voice/voice-dna.md`) — keeps the five hooks sounding like the author.
- Your banned-word / brand list (at She Leads AI that is `brand-enforcement/banned-items.md`, one file, includes the word swaps) — enforced on the hook and body.
- Your program / product descriptions (at She Leads AI that is `standards/programs/canonical-descriptions.md`) — when the post names one.
Never stop because an optional file is missing. Say which ones you did not have and run.
## Hook Rules — Built In
These hold regardless of what other files are present.
1. **Curiosity, not defense.** The hook opens a gap the reader wants closed. It never puts her on the back foot — no "you've been doing it wrong," no "the context you crammed in stopped working," no framing the reader's failure. Lead with insight, a specific, or the author's own position.
2. **No insult, no scarcity, no fear.** No "spots running out," "before it's gone," "last chance," "don't miss." No security or FOMO framing. The reader is already doing the work; talk to her that way.
3. **No logistics first.** A date, a time, a Zoom link, or "we invite you to" is never the first line. Every logistics-first post in the baseline is in the bottom third.
4. **Short enough to survive the cut.** In the current baseline the median hook is about 84 characters in both the strong and weak halves, so length alone does not predict performance. What does — anything over roughly 100 characters gets truncated before the idea lands. Over 100 needs a reason.
5. **Don't explain the post in the hook.** The hook earns the click. The body explains.
6. **Nothing invented.** Any fact not in the post body stays in brackets — `[the number]`, `[what she said]` — for the author to fill.
7. **Always offer options.** Never hand back one hook. Five, each a different angle, one recommended.
## Inputs
- The post body, or what the post is about
- The platform, since the truncation point differs
- Whether it has an image, video, document, or nothing
## Steps
1. Read the baseline, whatever optional context you have, and the post body.
2. **Who cares.** Run this first. Does the hook offer the reader something or announce something about the organization. Event logistics, scheduling, and "we invite you to" openers sit at the bottom of the baseline. If the hook fails this, say so and rework the premise instead of generating variants of a dead angle.
3. **Length.** Count the characters. Median in the baseline is about 84 either way, so judge length by whether the idea lands before the platform cut, not by a magic number. Anything over about 100 needs a reason.
4. **Truncation.** Say where the platform cuts and what the reader loses.
| Platform | Roughly cuts at |
|---|---|
| LinkedIn | 140 characters on desktop, less on mobile |
| Instagram | 125 characters |
| Facebook | 80 characters on mobile |
| Threads and X | The whole post is the hook |
5. **Baseline comparison.** Name which posts in `linkedin-baseline-30.md` this most resembles and how they did. Compare within format where possible.
6. **Hook Rules + brand pass.** Run the seven Hook Rules on the hook. If you have a banned-word list, run it on the hook and the body. No exemption here.
7. Write five hooks, each a different angle, each with its character count.
8. Recommend one and say why in a sentence.
9. Run the check-in questions.
## Angles
**Angles that have won on this list**
- **Give the thing away.** Publishing the offer in full, the price, the framework itself. The top post in the baseline by a factor of three.
- **Contrarian defense.** Pushing back on a claim about the audience. Drew the most comments in the set.
- **Flat opinion, unexplained.** "Life is too short for markdown files." Short, declarative, explains nothing in the first line.
- **First-person confession.** What the author did or refuses to do.
- **Personal news about a person.** Works here and does not work in an inbox. Keep it to people the audience knows.
**Angles that have lost on this list**
- Event logistics — a time, a Zoom link, a date as the opening line.
- "We invite you to" and other host-voice openers.
- Hooks over about 110 characters that explain the whole post before the cut.
## Output
```
HOOK AUDIT
Current — "[exact text]" ([n] chars, cuts after "[word]" on [platform])
Who cares PASS / FAIL — [one line]
Length PASS / FLAG — [n] chars, idea lands before the cut or not
Truncation PASS / FLAG — [what the reader loses]
Baseline [closest comparable posts and how they did]
Brand PASS / FLAG — [quote the violation, cite the rule]
FIVE HOOKS
1. [Angle] "[text]" ([n])
Why [one sentence]
2–5. [same shape]
RECOMMENDATION
[Which one, one sentence why.]
```
## Works With
Nothing here is required. The skill runs alone.
**Before this skill (if you have them)**
- A voice skill for the author — at She Leads AI, `slai-voice`. Run it first so the five hooks sound like the person posting.
- Whatever produced the post body — a promo skill, a content miner, a draft.
**After this skill (if you have them)**
- A brand / banned-word check — at She Leads AI, `brand-enforcement`. Full pass, no exemption.
- Your posting rail. Hand the finished post to however you publish.
**Sibling**
- `email-open-optimizer` — same method, different baseline, for the inbox. Never share lines between them.
**Standalone mode**
Paste a hook and get the audit plus five rewrites. Without the body, the who-cares check runs on the hook alone and says so.
## Gotchas
I know you'll want to reuse a strong email subject line as a hook, or the reverse — don't. The Pamela announcement proves the surfaces diverge. Re-score against the right baseline every time.
I know you'll want to apply the subject-line banned-items exemption here because this skill is the sibling of the one that has it — don't. A social post is public body copy. The list applies in full.
I know you'll want to explain the post in the first line so nobody is confused — don't. The two shortest opinion hooks in the baseline are both top performers and neither explains itself. The hook earns the click; the body does the explaining.
I know you'll want to open an event post with the date and time — don't. Every logistics-first post in the baseline is in the bottom third. Lead with why it is worth going, put the time in the body.
I know you'll want to cite engagement rate because AuthoredUp returns it — don't. That field came back above 100% for several posts and is not trustworthy here. Use impressions, reactions, and comments.
I know you'll want to invent a detail the author did not say — don't. Leave a bracket. This failed once in the sibling skill and the author caught it.
I know you'll want to treat a single outlier post as the pattern — don't. It is one post. Say "the top post did X" rather than implying every giveaway post will.
Add to this list as failures surface.
## Constraints
- Never move a line between this skill and `email-open-optimizer` without re-scoring.
- Never apply the subject-line exemption to a social post.
- Never cite engagement rate from this data set.
- Any fact not in the post body stays in brackets.
- Impressions are affected by format and by time since posting. Compare within format, and never rank a two-day-old post against a two-month-old one without saying so.
## Check-Ins
After the first run, ask
- Are you editing any of the five, and what are you changing?
- Did the recommended hook feel obvious, or did you reach past it?
Before wrapping up, ask
- Any failure pattern to add to gotchas while it's fresh?
- What else can I take off your plate?
## Changelog
| Date | Version | What Changed |
|------|---------|-------------|
| 08.17.26 | 1.1.0 | Self-contained. Hook Rules (curiosity not defense, no insult/scarcity, no logistics-first, options always) written into the skill instead of living in a memory file. Optional context no longer halts the run when missing. Baseline refreshed from AuthoredUp (Jun 29–Aug 14) and given a replace-with-yours header. `social-posting-engine` removed from the chain — retired. Which-skill-when block added. |
| 08.07.26 | 1.0.0 | Initial skill. Built as the social sibling of `email-open-optimizer`, scored on 30 LinkedIn posts pulled from AuthoredUp. Subject-line banned-items exemption deliberately not extended here. |
