Index of every page on this site: llms.txt
All skills
wdm0006 avatar

/marketing-pages

@909f594
by Will McGinniswdm0006/python-skills116 stars
18
Run it yourself

Designs landing, pricing, comparison and product pages that explain the product fast and convert honestly — one-sentence value proposition above the fold, a single primary call to action, real product screenshots instead of stock art, proof, a clear pricing table, fast responsive pages, and no dark patterns. Use when building or redesigning a landing page, marketing site, pricing or comparison page, app store–style product page, or reviewing why a product's public pages feel dated or don't convert.

  • 1 file
  • 10.3 KB
  • Updated last week
  • GitHub
Use this Skill: https://skilld.dev/gh/wdm0006/python-skills/marketing-pages

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈131 tokens for metadata: the name and description. ≈2.5k when used: this file.

Description uses 6.6% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈531 characters, or 6.6%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Designing Marketing Pages

A visitor decides in a few seconds whether a page is for them. Everything on a marketing page either answers "what is this, is it for me, what do I do next?" or gets in the way. Most dated landing pages fail on clarity, not style.

Contents

  • The page skeleton
  • Above the fold
  • Show the product
  • Proof and objections
  • Pricing pages
  • Comparison pages
  • Visual design
  • Performance and responsiveness
  • Honesty rules
  • Review checklist

The page skeleton

A dependable order for a product landing page:

  1. Hero — headline, one supporting sentence, primary CTA, product visual.
  2. The problem / who it's for — in the visitor's words.
  3. How it works — 3 steps or 3 key features, each with a real screenshot.
  4. Proof — numbers, logos, quotes, ratings (only real ones).
  5. Details for evaluators — integrations, privacy/security, platforms, FAQ.
  6. Pricing (or a link to it).
  7. Final CTA — repeat the primary action.

Above the fold

  • Headline: what the product does for whom, in plain words — "Book and track equipment rentals from your phone", not "Reimagine the rental experience".
  • Subhead: one sentence on the key differentiator or how it works.
  • One primary CTA with an outcome label ("Download for Mac", "Start free"); at most one secondary ("See how it works"). Three equal buttons means no choice.
  • A real product visual — a screenshot or short clip of the actual UI.
  • Price or "free" signal near the CTA when it removes a question.

Show the product

Real screenshots beat illustrations and stock photos: they prove the product exists and answer "what does it look like?". Use crisp, current captures with realistic (fake) data, cropped to the feature being described, framed consistently (same device frame or same shadow and radius). Retake them when the UI changes — stale screenshots are the most common accuracy bug on product sites.

Sample data is part of the claim

A screenshot can imply a result even when its caption makes no performance promise. In financial or quantitative products, extreme returns and near-perfect win records look like evidence of likely outcomes, even when the data is invented. A "sample data" label does not make an exaggerated result representative.

Define a screenshot-data brief before capturing: plausible values, a mix of outcomes, and any prohibited extremes. Use the same brief for new drafts and screenshots already on the public page or store listing. For example, show a portfolio with ordinary gains and losses rather than an unbroken series of wins; show the interface's ability to explain results rather than a spectacular result. Do not invent a universal "safe" return or win-rate cutoff: the sample must fit what the product can substantiate for its audience.

Keep synthetic demonstration data distinct from measured proof. If a figure is presented as an actual result, attach its evidence and evaluation protocol as below; otherwise make its illustrative role clear and avoid copy that treats it as a benchmark or expected outcome.

Proof and objections

  • Specific beats general: "Used by 1,200 teams" beats "Loved by teams".
  • Quotes with a name and role; ratings with their source.
  • An FAQ that answers the real objections: price, privacy, data export, platform support, what happens when you cancel.
  • Never fabricate testimonials, user counts, logos or ratings.

Keep measured proof attached to its protocol

A benchmark number is a claim about a particular evaluation, not a timeless property of the product. Before rewriting copy around new results, re-verify the claim evidence, re-run the proof measurement when a refresh is intended, then update the copy that consumes it. Writing copy first preserves the very numbers the refresh was meant to replace.

Put the identifying protocol beside each measured figure wherever it appears, not only in a linked report: package version, split versus walk-forward evaluation, warmup, and seed when applicable. For example:

Median processing time: 18 ms per item
Version 2.3; held-out split; 100 warmup items; seed 17.

Two pages quoting results from different protocols can look like a regression even when both numbers are correct. Link the full evidence from the local caption; a stable claim ID does not replace the context needed to interpret the number.

Pricing pages

  • 2–4 plans in columns; recommended plan visually marked; monthly/annual toggle that shows the real per-month price and the saving.
  • Each plan lists what is included in the same order, with the differences first.
  • The price shown is the price paid — taxes and limits stated plainly near it.
  • A free tier or trial says exactly what is free and what happens at the limit.
  • FAQ under the table for billing questions.

Comparison pages

"X vs Y" and "alternatives to Y" pages are high-intent and worth building: a fair feature table, honest about where the alternative is better, with a clear "choose us if…" / "choose them if…" section. Keep competitor facts dated and sourced; inaccurate comparisons damage trust and can create legal exposure.

Visual design

The designing-interfaces skill's fundamentals apply, tuned for reading:

  • Larger type than the app: 18–20px body, 40–64px hero headline on desktop, scaling down with clamp() on phones.
  • Generous vertical rhythm between sections (64–128px); one idea per section.
  • Prose blocks at ~65 characters wide even inside wide layouts.
  • One accent color reserved for CTAs so the action is always the most visible thing.
  • Consistent section structure (heading, one sentence, visual) so the page scans.

Performance and responsiveness

Marketing pages are judged on first load, often on a phone:

  • Largest Contentful Paint under ~2.5s: compress and size images (srcset, WebP/AVIF, explicit width/height to avoid layout shift), lazy-load below the fold, avoid render-blocking scripts and heavy font stacks (one family, two weights, font-display: swap).
  • Check with npx --yes lighthouse "$URL" --only-categories=performance,accessibility,seo --view.
  • Design the phone layout first; stack sections, keep the CTA reachable, and make sure screenshots remain legible when scaled down (crop tighter for mobile).
  • Meta title, description and an Open Graph image so shared links look intentional.

Honesty rules

No dark patterns: no fake countdown timers or scarcity, no pre-checked upsells, no hidden costs revealed at checkout, no confirmshaming ("No thanks, I don't like saving money"), cancellation as easy as signup. Every claim on the page must be true of the shipped product today — check copy against the live app before publishing.

Separate development, distribution, and claimability

A feature on the default branch is not evidence that a visitor can use it. Before updating copy or replacing screenshots, record three separate facts:

State Evidence to check
Implemented The feature exists in the current source tree.
Distributed The version/build available through the page's download or store CTA includes it.
Claimable today That distributed build exposes it to the audience and platform the claim addresses.

For example, a redesigned dashboard may be merged while the store still serves an older interface. Keep the current page's screenshots aligned with the store build; retain captures of the redesign as unpublished launch assets. They are premature for current copy, not stale assets to discard. Even inclusion in a build is insufficient if the advertised audience cannot access the feature.

Record the checked version/build and verification date with the claim evidence. Recheck distribution before publishing a prepared launch draft; a source merge alone must not promote future copy into present-tense claims.

Keep stable claims separate from dated observations

A claim ledger lets copy cite verified facts without duplicating their evidence. Give each reusable claim a stable identifier, record its evidence in the ledger, and cite that identifier from the copy. Append new identifiers; never renumber or reuse one that existing content cites.

Keep facts that change independently of a code change in a separate, dated observation section: store ratings, the latest released version and release date, and listing category order. Do not give these observations evergreen claim IDs. A citation preserves provenance; it does not keep the fact current.

Claim ledger
C1: Data is stored on the device. Evidence: verified storage behavior of build 7.

Dated observations (not reusable claim evidence)
2026-06-12: Store listing shows version 2.4 and a 4.6 rating.

If a page needs a volatile fact, verify it again for publication and present it with an as-of date and source, or omit it. Do not promote an observation into a stable claim merely to satisfy a copy citation check. Keeping the two sections separate makes the next refresh explicit instead of letting an old observation survive indefinitely behind a valid-looking citation.

Review checklist

  • A stranger can say what it is and who it's for after 5 seconds above the fold
  • One primary CTA, repeated at the end, labelled with the outcome
  • Real, current product screenshots; sample values are representative, including on existing public listings
  • Proof is real and specific; FAQ covers price, privacy, export, cancellation
  • Measured proof refreshed before dependent copy; evaluation protocol shown beside each figure
  • Pricing shows the real price and limits
  • Phone layout checked; Lighthouse performance and accessibility reviewed
  • Every claim and screenshot verified against the build available through the CTA, for its advertised audience
  • Reusable claims have stable evidence IDs; volatile facts are rechecked and dated, not cited as evergreen claims

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 909f594. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week

README badge

README badge for wdm0006/python-skills/marketing-pages