sidchaudhary/gtm-skills / skills/brand-designer/hook-writer
hook-writer
VerifiedTakes a LinkedIn post draft and generates four scored hook variants for the opening one or two lines before the see-more cutoff, each built on a distinct psychological framework, with a recommendation and the reason it wins. Use when a post's opening feels flat, or when testing hooks before publishing. Boundary: LinkedIn post openings only. `cold-email` does the equivalent job for cold email subject lines, and `cold-email` writes the opening line of a cold email body.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.4 KB |
| references/house-rules.md | markdown | 13.3 KB |
| references/outbound-copy-standards.md | markdown | 15.5 KB |
references/outbound-copy-standards.md
View on GitHub# Outbound Copy Standards
The shared bar for every line of copy this pack produces that a prospect or customer will
actually read: cold emails, subject lines, LinkedIn posts, SMS and push, landing page copy, save
offers, one-pagers, objection responses, and video on-screen text.
Two things separate this from a general writing-style guide. First, it is calibrated to **where
the reader sits on the awareness ladder**, because the same sentence that converts a
product-aware buyer reads as noise to someone who does not yet know they have the problem.
Second, its failure list is drawn from **what kills reply and conversion rates**, not from what
makes prose inelegant. Some of the worst-performing outbound is competently written.
Read this before writing. Check against it before returning.
---
## 1. Awareness stage decides the shape of the message
Diagnose the stage before writing a word. If the inputs do not tell you, ask: "does this person
already know they have this problem, and do they know solutions like ours exist?" Guessing the
stage is the single most common reason good copy lands badly.
| Stage | What the reader already believes | What the copy must do | What kills it |
|---|---|---|---|
| **Unaware** | Nothing is wrong | Name the problem through a consequence they have felt. No product. | Leading with the product. Reads as spam because there is no problem in their head to attach it to. |
| **Problem aware** | Something is wrong, unsure what | Define the problem precisely and price the cost of staying stuck | Jumping to features. They cannot evaluate a solution to a problem they cannot yet articulate. |
| **Solution aware** | Solutions exist, comparing approaches | Explain why this *approach* fits their situation, not why this vendor is good | Vendor-vs-vendor comparison before the approach is settled |
| **Product aware** | Knows this product, hesitating | Remove the specific hesitation with proof, differentiation, specificity | Re-explaining the problem. Condescending, and wastes the one thing they came for. |
| **Most aware** | Ready, needs a reason now | Make the next step frictionless and give a real reason to act today | Manufactured urgency. At this stage a fake deadline is the fastest way to lose a ready buyer. |
**Trigger signals usually reveal the stage.** An open req for a role that owns the problem is
problem-aware or later. A competitor's tool in the stack is solution-aware. A pricing-page visit
is product-aware. A funding round tells you about budget, not awareness, so it does not move
the reader up the ladder on its own.
**Stage mismatch is a rewrite, not a polish.** If the draft addresses the wrong stage, changing
the words will not save it. Change the argument.
---
## 2. Promise continuity
Every attention-getting element makes a promise. The body has to pay it.
- A subject line's promise must be resolved by the email it opens. If the subject says
"three tools, one funnel" the body has to be about that, in the first sentence.
- A hook's promise must be resolved by the post under it.
- An ad or headline promise must be resolved by the page it lands on.
When a skill receives both the attention element and the body, **verify the promise against the
actual body text rather than assuming it holds**. A subject line that outperforms on opens and
underdelivers in the body trains the reader to stop opening. That cost lands on every future
send, not this one, which is why it rarely shows up in the metric being optimized.
If the promise cannot be paid, change the promise. Do not weaken the body to match a subject
line you liked.
---
## 3. The opening line
The opening is the whole message in outbound. It gets one job: prove this was written for this
person and not to a list.
**A real opening references something checkable.** A named trigger, a specific line they wrote,
a decision they made, a number from their own world. Something they could confirm.
**Fake personalization is worse than none.** These read as automated even when a human wrote
them, because the reader has seen the template:
| Tell | Why it fails |
|---|---|
| "I noticed you're the [title] at [company]" | Merge-field personalization. Proves only that their title was in a database. |
| "I came across your profile and was impressed by..." | Flattery with no object. Works identically for anyone. |
| "As a [role], you probably struggle with [generic pain]" | Assigns them a problem instead of observing one. Invites "actually, no." |
| "Hope this finds you well" / "Hope you're having a great week" | Costs a line, says nothing, signals template. |
| "Quick question" as an opener or subject | The most-used opener in B2B outbound. Signals volume sending. |
| "I'll keep this short" | Then be short. Saying it is the opposite of it. |
| **Bio recap**, describing their company back to them from their own About page | They wrote it. Repeating it proves you read a page, not that you understood a situation. This is the most common "personalised" opener that still reads as automated. |
| **Compliment with no bridge**, "I love what you're building" | Praise that leads nowhere. An observation has to imply a consequence, or it is filler with a nicer tone. |
| Congratulating a funding round with no read on what it changes | Everyone sent that email the same week. |
**The specificity test:** could this exact opening line be sent unchanged to a different person
at a different company? If yes, it is not personalization. Rewrite it.
---
## 4. Proof
Vague outcome claims are treated as noise by anyone who reads outbound regularly. Proof has a
ladder, and every rung down costs credibility:
1. **Named customer plus a number.** "Fernbrook cut setup from six weeks to nine days."
2. **A number with a described customer.** "A bootstrapped DTC brand cut it from six weeks to nine days."
3. **A number alone.** "Teams cut setup by 80%."
4. **A described mechanism, no number.** "Setup skips the reconciliation step entirely."
5. **An adjective.** "Significant time savings." This is not proof. Do not ship it.
Use the highest rung the user actually supplied. **Never move up the ladder by inventing the
missing part**, do not attach a customer name to an unattributed stat, do not invent a
percentage for a mechanism, and do not convert "faster" into a number. If the user only gave you
rung 4, ship rung 4 and tell them what rung 1 would need.
A number the reader cannot connect to their own situation is weaker than a mechanism they can.
"40% lift" from a company unlike theirs loses to "you stop reconciling three dashboards."
---
## 5. One ask
Every message gets exactly one ask, and the ask has to be smaller than the reader's current level
of trust.
- **Cold, first touch:** ask for interest, not time. "Worth a look?" beats "do you have 20
minutes Thursday?" A calendar request assumes a decision they have not made.
- **Warm, replied once:** ask for time, and name it precisely. Two options beat an open question.
- **Never stack asks.** A question plus a link plus a PS with a second option is three asks. The
reader picks none.
- **A question mark is not automatically an ask.** "Thoughts?" is not a next step.
---
## 6. Slop that specifically costs replies
General writing guides catch most prose slop. These are the ones that survive an ordinary edit
and still lose deals:
**Feature listing where an outcome belongs.** "We offer analytics, journeys, and attribution" is
a menu. The reader has to do the translation work into their own problem, and they will not.
**Hedged claims.** "Can help teams potentially reduce..." Every qualifier is read as an admission
that it does not work. Either claim it or cut it.
**Borrowed urgency.** "Prices go up Friday" when nothing about their situation changes Friday.
Real urgency comes from *their* clock: a role opening, a renewal, a launch, a season.
**Us-first sentence order.** Count the sentences that start with "we" or the product name against
the ones that start with the reader's situation. In cold copy, the reader's world comes first.
**Insider vocabulary as a shortcut.** Category language ("agentic GTM", "composable CDP") assumes
the reader shares your frame. At unaware and problem-aware stages they do not, and the term does
the opposite of clarifying.
**Length used as effort.** A long message does not read as more thoughtful, it reads as more
expensive to answer. If a sentence does not change whether they reply, cut it.
**Em dashes as rhythm.** In short copy use none. They cluster, and clustered dashes are one of
the most reliable machine-written tells in email.
**Formatting as emphasis.** Bold inside an email body, bullets in a five-sentence message, and
emoji in a subject line all read as marketing rather than a person writing.
---
## 6a. The cold first-touch structure
The one-ask rule and the opening-line rules above say what a cold email must not do. This is the
shape that works when it is doing it right, and it is tighter than most senders expect.
**Word budget: 55-90 words for a first touch.** Under 120 is a ceiling for the format, not a target.
A first cold email at 110 words is usually three sentences of setup that a second draft removes. The
shorter version reads as a colleague noticing something; the longer one reads as a campaign.
**Five parts, in order, one to two sentences each:**
1. **The observation.** One specific, checkable fact about their situation, with the pain implied
rather than stated. Max 20 words. Something they could confirm, and that would break if sent to
anyone else.
2. **Who you are, compressed.** "I'm [name] at [company]." One line. Not "My name is X and I work at
a company called Y, which helps teams to…", a long self-introduction spends the reader's attention
on the sender at the exact moment it is worth the least.
3. **What the product *is*.** Plain language, in one sentence. Not what it will do for them: that is
the next line's job, and collapsing the two produces a claim before they know what is being
claimed about.
4. **The bridge.** One sentence connecting the product to *their* specific situation, referencing the
observation from part 1. If this sentence would read identically to a different prospect, part 1
was not specific enough.
5. **One ask.** One sentence, and shorter than feels comfortable.
**Imply the pain, do not assign it.** "Running lifecycle for six clients across three separate
accounts" lets the reader supply the problem themselves. "You're probably struggling to keep those
accounts in sync" tells them what their problem is, invites "actually, no", and reads as a script.
### The ask, specifically
Short beats warm. These soft closes all read as a script and all reduce replies:
- "Would love to hop on a quick call"
- "Happy to share more if useful"
- "Let me know if this resonates"
- "Thoughts?"
- "Open to a conversation?"
They are hedges: they ask for permission to ask. A direct short question ("Worth a look?", "Want
in?") outperforms them because it is answerable in one word.
### If the offer involves scarcity, name the number
"One of about 15 lifecycle consultants we reached out to" is checkable and therefore credible. "A
select group", "a few people", "a handful of teams" are unfalsifiable and read as sales language. If
the real number cannot be stated, drop the scarcity framing rather than vaguen it, see the
borrowed-urgency rule above.
### Subject lines for a cold first touch
The subject-line rules in `references/email-templates.md` cover campaigns. Cold first touches behave
differently, and the pattern that works is narrower:
- **Two to four words. Lowercase. No punctuation** beyond a comma between two proper nouns. It should
look like something a person typed in a hurry to one recipient, not something a marketing tool
produced.
- **Draw on one of these, in priority order:** a trigger event ("just went independent"), an
information gap the reader will want closed ("klaviyo to what exactly"), a named transition
("gymshark to adanola"), or an observable fact about how they operate ("fractional, six clients").
- **Do not use** the recipient's first name alone, their city, your own product or company name, or
any of: free, opportunity, demo, growth, scale, strategy, automation, funnel, campaigns, lifecycle,
email. These are category words. They tell the reader which folder the message belongs in before
they open it.
## 7. Voice
If `.agents/product-context.md` exists, its brand voice and banned-word list are **binding, not
advisory**. A banned word in returned copy is a failed check, not a stylistic preference.
Where no voice profile exists, default to plain operator register: short sentences, concrete
nouns, active voice, no category jargon, and no enthusiasm the reader has not earned yet.
---
## 8. Check before returning
Numbered so a specific check can fail. If any fails, fix it and re-check. Do not return copy
that fails a check, and do not note the failure as a caveat instead of fixing it.
**Stage and argument**
1. Is the awareness stage stated or inferable from the inputs, and does the argument match it?
2. Does the copy avoid pitching a solution to a reader who cannot yet name the problem?
**Promise**
3. Does every subject line, hook, or headline get paid off by the body it introduces?
4. Was that verified against the actual body text, not assumed?
**Opening**
5. Does the opening reference something specific and checkable about this reader?
6. Would the opening survive the specificity test, would it break if sent to someone else?
7. Is every tell from the fake-personalization table absent?
**Proof**
8. Is every proof point at the highest ladder rung the user actually supplied?
9. Was nothing invented to climb the ladder: no attached customer name, no fabricated number?
**Ask**
10. Is there exactly one ask, sized to the reader's current trust?
10a. For a cold first touch: is the body 55-90 words rather than merely under the 120 ceiling?
10b. Does it follow the five-part shape, observation, compressed self-intro, what the product *is*,
the bridge to their situation, one ask, with none of the five collapsed into another?
10c. Does the bridge sentence reference the observation specifically enough that it would not read
identically to a different prospect?
10d. Is the pain implied rather than assigned, so the reader supplies it instead of being told what
their problem is?
10e. Is the ask a short answerable question rather than one of the hedged soft closes ("hop on a quick
call", "if this resonates", "happy to share more", "thoughts?")?
10f. If scarcity appears, is a real number named rather than "a select group"?
10g. For a cold subject line: two to four words, lowercase, no punctuation beyond a comma between
proper nouns, drawn from a trigger / information gap / named transition / observable fact, and
free of category words (free, opportunity, demo, growth, scale, strategy, automation, funnel,
campaigns, lifecycle, email) and of the sender's own company name?
**Slop**
11. Are feature lists replaced by outcomes, hedges cut, and borrowed urgency removed?
12. Do more sentences start with the reader's situation than with "we" or the product name?
13. Is category jargon absent unless the reader is product-aware or most-aware?
14. Em dashes: none in short copy, and no clusters anywhere?
15. Is formatting plain, no mid-sentence bold, no bullets in short messages, no emoji in subject lines?
**Voice**
16. Does the copy contain zero words from the banned-word list in `.agents/product-context.md`?
17. Does it read as the brand's register rather than generic marketing voice?
**Honesty**
18. Is every company-specific detail traceable to something the user supplied, with nothing
filled in from memory or plausibility?
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |