Write a sales page — one long-scroll page that sells ONE thing. Titles first (the full title chain designed before any body), then the sections under it. TRIGGER when a page's job is to close one offer in a single scroll. DO NOT TRIGGER for a page that trades for an email (write-optin-page), a page that lists several offers or pages (write-catalog-page), the spoken script behind a demo (write-vsl), or a standalone headline or lead (write-headlines, write-leads).
Scanned 9/5/2026
Install to Claude Code
npx -y skills add heyJordanParker/dotfiles --skill write-sales-page --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Write Sales Page?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/heyjordanparker-write-sales-page)More formats (shields.io, HTML) on the badges page.
---
name: write-sales-page
description: Write a sales page — one long-scroll page that sells ONE thing. Titles first (the full title chain designed before any body), then the sections under it. TRIGGER when a page's job is to close one offer in a single scroll. DO NOT TRIGGER for a page that trades for an email (write-optin-page), a page that lists several offers or pages (write-catalog-page), the spoken script behind a demo (write-vsl), or a standalone headline or lead (write-headlines, write-leads).
---
# Write Sales Page
One Process: design the full title chain first, because skimmers read only titles — the chain alone must tell the whole story of selling this one thing; then write each section under it.
Write under the plan's Reader.md, Brief.md, and Proof.md, and the product's Voice.md, which names the WHO speaking (founder, company, or champion); the whole page speaks in that one voice. The offer, its mechanism, and its proof come settled from the owner (draft-offers, mechanism); never assume an offer mechanism (trial, guarantee, urgency, anchor price) the owner has not approved, and never invent a number, quote, or customer.
## 1. Design the title chain before any body
The chain answers the reader's ordered BUYING QUESTIONS — real questions born of not understanding the product, in journey order — and marches as one argument to the CTA.
### Default to the structure readers expect
The chain and section order follow how a sales page is usually presented; creativity lives in the words, not the skeleton.
Evidence (owner, verbatim): "people are USED TO certain designs & us being creative in that sort of structure is almost always a terrible idea. we CAN use this as a surprise or to subvert expectations but in this case (& MOST cases) this is just a sloppy structure"
The sales-page skeleton (pick the beats the argument needs, in this order): hero (promise + subhead at the reader's stage) → picture-the-after (the vivid dream outcome in words, never dependent on an image) → problem agitated (end on the reader's limiting belief) → mechanism reveal (the named mechanism, honest because real) → solution as the mechanism (show the access shape) → what it does for them (outcome-first; features appear as a features section for ultra-aware readers, per copywriting's features law, which also holds the owner-open benefit form) → proof after each benefit and beside each CTA → price anchored to a REAL number (a competitor's cost, or the cost of inaction — never an invented struck-through price) → offer + price box + outcome-named CTA → risk reversal at the CTA → value framing anchored to the real alternative cost (not a bonus stack with invented retail values) → why-now that is real → FAQ → closing CTA (recap the promise, repeat the ask, re-attach the risk reversal) → legal footer.
### Write the chain as one argument marching to the CTA
Each title is a step the next advances — a problem title ANSWERED by a later solution title, closing on a resolution CTA title. A chain that skims as a list of topics fails even when every title passes alone.
### Head product sections with claim titles, proof and FAQ with carried questions
Claim titles are standalone, falsifiable, fifth-grade, and dead when a competitor's name is swapped in (the platitude test). Proof, trust, and FAQ sections take the reader's own first-person carried question, never a page-voiced "you" question. Never quiz or sentence the reader ("Do you actually know…", "What your setup can't show you is…"). An unrecognized problem is a body argument, never a heading.
### Run the mechanical checks on every title
No pronoun leaning on anything above ("it", "these numbers"); no page-geometry deixis ("the section below"); every title true and complete out of context; max two consecutive titles on the same opening subject; never raise an objection the reader does not carry yet.
If the chain departs in a big way from the expected sales-page structure, ship the departure only as a stated decision with its strategic reasoning attached.
## 2. Model the reader at each scroll position, then write the section
Three lines before writing each section: what they now believe, the loudest doubt, what they would leave over. The section contains only what moves that doubt and ENDS when the doubt is answered.
### Hold one Reader through the page
Every section assumes the same Reader — Reader.md's Audience, the named buyer — and each section answers the question the previous raised in that Reader. Never order sections by a teaching sequence instead of the chain of questions one held Reader actually asks.
### Name the Reader's motivation before writing any answer
Before an answer, name WHY the Reader asks; the answer serves that motivation and adds information beyond the question.
Evidence (owner, verbatim): "people asking this question need REASSURANCE. the anser gives none of that. (you need to ask WHY are people doing things & deeply understand EVERY action's motivation to write good copy)"
### One point per section
Never mix angles; pick the stronger and cut the other. People came for the software — side topics (the founder's story) get a sentence and a half.
### Answer result-first
A yes-question opens with "Yes" and the result the reader wants; mechanics get almost no space; developer language never reaches a non-developer reader.
### Carry the one belief shift as a story
At most one belief-shifting argument per page, and it arrives as a story, never a lesson (write-story produces the narrative component).
## 3. Draft as speech, then compress to sales register
Speak each section to one skeptical equal, transcribe, then compress hard: skimmable, tight, the conclusion arrives fast. No pseudo-conversation filler, no essays, no explanation-of-the-explanation, no teaching.
## 4. Hold the language bar
### Write fifth-grade words
A ten-year-old parses every sentence on one read. No cuteness, no coined nouns, no platitudes, no prose math (worked figures are design elements).
### Ban metaphors and metaphorical verbs
A verb states the literal action; a verb that imports an image makes the reader stop and interpret.
Evidence (owner, verbatim): "avoid metaphors AND metaphorical verbs; AIs tend to overuse BOTH in absurd quantities and it creates way too much unnecessary thinking & interpretation for the text"
### Be specific
Every claim, example, and figure is the specific one; a line that gestures at a category instead of naming the thing fails.
Evidence (owner, verbatim): "be specific. good copy IS specific."
### Punctuate for consumption, not grammar
Punctuation serves the reader's eye; drop grammatically correct marks that slow the read. No em dashes, no colon-drama, no contrast ticks ("X, not Y").
Evidence (owner, verbatim): "Copywriting is about making text EASY to consume, not adding every grammatically correct comma."
### Never narrate the reader
No "you"-assertion of what they do, have, feel, or pay. Reference the reader only through a first-person carried question, a subordinate time clause of visit-proven behavior, a generic subject, or the approved possessives. Never enumerate the reader's assumed app stack — one wrong guess poisons the page.
Late SaaS objections shape: [references/faq.md](references/faq.md).
## 5. Express visuals as design requirements
Enumerations of kinds, industries, or examples — and any content a visual carries better than prose (plans, worked figures, capability grids, pricing) — are a design requirement: the named visual plus the copy strings on it, never a prose list. Plans and pricing are shown, never summarized in prose. Output home: the deliverable's Wireframe.md, in the section's layout block.
Every section also carries a layout block — its section shape, where each element sits, and its CRO notes — written into Wireframe.md per the wireframe format in [references/wireframe.md](references/wireframe.md). Copy owns where things go and why; design owns how they look.
Evidence (owner, verbatim): "We need our copy team to be able to express requirements for design/images; otherwise they do stupid shit like that. No one is reading a paragraph listing a dozen industries."
## 6. Close forward
The close names the forward step — never rescuing the scene's own past artifact, never the hero's phrases, never the reader's failure as punchline, never a quotable clincher. The product's differentiation is spent exactly once on the page, by its owner section. Pain appears only where the product reduces it.
## 7. Wireframe the converged copy
Once the copy is locally clean, wireframe the finished page into the deliverable's Wireframe.md — section order, per-section rationale, a layout block per section, and the title chain — per the format in [references/wireframe.md](references/wireframe.md). This is a production output the writer owns; design works from it.
Verification: the bare title chain reads top to bottom as one argument selling one thing; every title passes the mechanical checks; the draft cites no unpermitted fact and contains zero banned strings; every section reads aloud as a person talking — before it reaches the gate.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!