agentleFS
fortunto2fortunto2/solo-factory/skills/landing-gen/SKILL.md

solo-landing-gen

Skill18 starsChanged 4 months agoMITView on GitHub

Turn a PRD into a landing page that converts and ships honest. One batch of intake questions, section-by-section copy with A/B headlines, a type, spacing, radius and motion system on Tailwind steps, a grep gate, then SEO with JSON-LD, agent-readiness files and a cookieless analytics counter. Use when user says "create landing page", "сделай лендинг", "landing page design", "сверстай лендинг", "лендинг для приложения", "write landing copy", "hero section", "A/B headlines", "landing content", or needs a marketing page from a PRD. Do NOT use for auditing a live site's SEO (use /seo-audit), social media posts (use /content-gen), or dense B2B app screens and dashboards.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Copy reads like a person: No em dashes, no stock inflation, specific over vague, measured over specific (/humanize).

    line 69
  2. Distinctive:

    Values come from tokens: Every size, gap and radius is a scale step.

    line 74
  3. Very specific:

    Fix: Run /validate to generate the PRD first, or provide a README with problem, solution, features.

    line 141
  4. Very specific:

    Fix: Add research.md with competitive analysis and real user quotes.

    line 145
  5. Distinctive:

    Fix: Put it in the tokens block as a named value and use the token.

    line 153

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/landing-gen

One page, one offer, one audience, one primary action. This skill takes a project's PRD to a landing page that says one thing, is set on a system rather than by eye, and ships with the parts that usually get forgotten: honest proof, JSON-LD, the files agents read, and a counter that needs no cookie banner. With astro-static or Next.js it builds the page; otherwise it writes the content file.

Stages

Load only the reference for the stage you are in. Each opens with When / Inputs / Output.

#StageOutputRead
1IntakeBrief: action, offer, ICP, traffic source, objections, proof on hand, layout type A-Ereferences/intake.md
2CopyOutline, hero, 5 A/B headlines, benefits, how it works, proof, FAQ, final CTAreferences/copy-sections.md
3Visual systemOne tokens block (font, type scale, spacing, radius, surfaces, motion), then sections one by onereferences/layout-type-spacing.md
4ShipHead and OG, JSON-LD with sourced Offer, robots/llms.txt/404, counter and CTA events, legal, formreferences/ship.md
5CheckGrep gate prints nothing, checklist ticked, /humanize, seo audit, seo agent-auditreferences/anti-patterns.md

MCP tools (use if available)

  • kb_search(query): related methodology (conversion, copywriting)
  • project_info(name): project stack and details
  • web_search(query): competitor landing pages

Without MCP, fall back to Glob + Grep + Read.

Steps

  1. Parse the project from $ARGUMENTS. Read docs/prd.md, README.md, docs/research.md. Nothing there: run /validate first, or take the product description from the user.
  2. Detect the stack. astro.config.* → astro-static: src/pages/index.astro plus one component per section. next.config.* → a Next.js route. Neither: content only.
  3. Intake (intake.md). Fill the brief from the documents, ask only the gaps in ONE AskUserQuestion batch, mark unanswered fields (assumed) and keep going. Pick the layout type and the mode: fake-door (hero, CTA, email capture, about two hours) or the full page.
  4. Forced reasoning, written into the brief before any headline: the ONE conversion, each top objection with the proof or reversal that answers it, what sits above the fold.
  5. Competitor landings (optional, with search available): top three pages, their headline patterns, CTA wording, proof types, and the positioning gap they leave.
  6. Copy (copy-sections.md). The whole outline and all copy go into docs/landing-content.md before any markup.
  7. Build (layout-type-spacing.md). Tokens first, then hero, benefits, how it works, proof, FAQ, final CTA, one section per iteration. Never regenerate the whole page to change a part.
  8. Ship (ship.md): head, JSON-LD, agent files, counter with CTA events, footer, form.
  9. Check (anti-patterns.md): the gate must print nothing; /humanize the copy; after deploy, seo audit and seo agent-audit on the live URL (/seo-cli).
  10. Summary: headline, CTA, layout type, assumptions made, gate result, next steps.

Rules that hold in every stage

  • One action. A second CTA is a subordinate link, never an equal button.
  • Proof is real or absent. Invented quotes, logos, counts or ratings never reach a page that ships; sample content is marked [SAMPLE] and the gate fails on it.
  • Copy reads like a person. No em dashes, no stock inflation, specific over vague, measured over specific (/humanize).
  • Privacy-first. A cookieless counter (superduper-analytics by default, never PostHog or GA), self-hosted fonts, no third-party pixels. Then no cookie banner is needed and the privacy page can say so (/legal).
  • Values come from tokens. Every size, gap and radius is a scale step; an arbitrary value is a gate failure, not a style choice.

Output: docs/landing-content.md

# Landing Page Content: {Project Name}

**Generated:** {YYYY-MM-DD}
**Primary action:** {action} · **ICP:** {persona} · **Traffic:** {source}
**Layout:** {A-E}, because {one line} · **Mode:** {fake-door | full}
**Assumed:** {fields marked (assumed), or "none"}

## Brief
Objections → answers: {objection → proof / reversal / FAQ}

## Hero
**Headline:** {headline}
**Subheadline:** {subheadline}
**CTA:** {button text} · **Proof signal:** {real item or "none yet"}
**Visual:** {what is shown}

## A/B Headlines
1. {pain} 2. {benefit} 3. {curiosity} 4. {social proof, real N only} 5. {direct}

## Benefits
## How it works
## Proof
## FAQ
## Risk reversal
## Final CTA

## SEO and JSON-LD
{title, description, index/noindex, JSON-LD types and Offer source}

## Analytics events
{event names and where they fire}

---
*Generated by /landing-gen. Replace every [SAMPLE], run the gate, then publish.*

Notes

  • Fake-door test: hero, CTA and email capture are enough; noindex it.
  • Headlines are testable: run two or three variants, not one.
  • Works with /seo-audit and /seo-cli: generate, deploy, then audit the live page.

Gotchas

  1. Hero CTA must be above the fold — if users need to scroll to find the primary action, conversion drops 40%+. Hero = headline + subheadline + CTA button + visual. Nothing else.
  2. Too many CTAs = decision paralysis — one primary CTA per page. Secondary actions (learn more, see pricing) should be visually subordinate. Never put 3 equal-weight buttons side by side.
  3. Social proof placeholders hurt more than no social proof — "Join 0 users" or empty testimonial slots signal "nobody uses this". Remove social proof section entirely until you have real data. Use "Built by {founder name}" instead.
  4. OG image is required for social sharing — without og:image, links shared on Twitter/LinkedIn/Slack show a blank preview. Use 1200x630px. Test with https://cards-dev.twitter.com/validator.
  5. A/B test headlines BEFORE launch — don't pick one headline and ship. Use Google Ads ($20/day for 3 days) to test 3-5 variants against each other. The winner goes on the landing page.
  6. Fonts from Google's CDN leak every visitor's IP. A German court awarded damages for it (LG München I, 3 O 17493/20, January 2022). Self-host: Fontsource in Astro, next/font in Next.js.
  7. A reveal animation can hide the page. Content set to opacity: 0 in CSS stays invisible when JS fails or is off. Hide only under a .js class set inline in <head>.
  8. An Offer with no source is a claim agents repeat. Price and rating in JSON-LD come from the store listing, a registry or the page itself, or they are left out.

Common Issues

No PRD or product info found

Cause: Project lacks docs/prd.md or a README with a product description. Fix: Run /validate to generate the PRD first, or provide a README with problem, solution, features.

Headlines too generic

Cause: Weak problem statement or no competitor differentiation. Fix: Add research.md with competitive analysis and real user quotes. Specific pain produces specific headlines.

Stack not detected for page scaffolding

Cause: No astro.config.* or next.config.* found. Fix: The skill writes content-only markdown. To get page files, use the astro-static or nextjs-supabase stack.

The gate flags a value the design really needs

Cause: A one-off size, such as a hero illustration's exact width. Fix: Put it in the tokens block as a named value and use the token; the gate checks classes, so a named token passes and stays findable.

Credits

Intake structure, the A-D layout taxonomy, page structure, type-scale snapping, the spacing table, the nested-radius rule, motion and state rules, and the checklist shape were adapted from landing-page-design by Elaya (MIT, Copyright (c) 2026 Elaya), rewritten for this skill's conventions. See THIRD-PARTY.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.