Landing Page Creation

Landing-page creation

You are building a marketing landing page for a SeldonFrame workspace. The workspace already exists (created by create_workspace_from_url), already has a CRM, booking page, intake form, and AI chatbot. Today its public surface is a chatbot-only preview page. The operator has asked you to replace that with a full marketing landing.

Mental model

SeldonFrame's pitch is "ops stack first, marketing optional." The operator already has the ops stack live. A landing page is opt-in — some clients have great existing sites and only need the chatbot+CRM bolted on; others have outdated sites and want SF to render a new one.

When you build a landing page, the operator is in the second bucket. Your job is to produce a page that's better than what their client currently has — premium, niche-specific, archetype-correct.

Process

Step 1 — Understand the workspace

Call get_workspace_state(workspace_id) to load:

• The soul (business name, vertical, services, voice, certifications, reviews, emergency-service flag, same-day flag, business description)
• The classified aesthetic_archetype (one of 7: bold-urgency, clinical-trust, cinematic-aspirational, editorial-warm, technical-restrained, soft-residential, brutalist)
• The theme (palette, fonts) — already archetype-correct from v1.40
• Active integrations (so you don't reference an API the workspace doesn't have configured yet — e.g. don't promise SMS booking confirmations in copy if Twilio isn't connected)

If aesthetic_archetype is null (pre-v1.54 workspace), pick one yourself from the vertical + business description, then proceed.

Step 2 — Optionally consult external design skills

If the user has installed claude-code/frontend-design/SKILL.md (the Anthropic frontend-design plugin), read it for Tailwind / Framer Motion patterns and component composition guidance. It pairs with this skill: frontend-design gives you the HOW (components, motion, layout); this skill gives you the WHAT (which SF blocks, what order, what voice).

If google-labs-code/design.md is available, optionally invoke it for design-language generation. It produces a design.md you can fold into the block prompts as additional constraints.

Step 3 — Decide the block sequence

Pick from available blocks. The default sequence for most service businesses:

1. hero — primary headline + CTA + supporting visual
2. servicesGrid — what they do (cards, pricing optional)
3. (optional) projectGallery — visual proof of work
4. (optional) testimonials — social proof
5. faq — top objections answered
6. (optional) emergencyStrip — only for bold-urgency archetypes
7. cta — closing call-to-action

Per-archetype adjustments

Step 4 — Generate and persist each block

For each block in the sequence:

a. Call get_block_skill(block_name) — returns the block's SKILL.md with its prop schema, voice rules, validators, and worked examples.

b. Generate props following the SKILL.md AND the archetype voice (leanInto / avoid lists from the archetype registry — accessed via the v1.54 theme.aestheticArchetype value).

c. Call persist_block(workspace_id, block_name, prompt, props) — v1.54's server enforcement WILL override your template and variant fields if they don't match the archetype's defaults. This is intentional — trust the server-side enforcement. Don't second- guess archetype defaults.

d. Note validator warnings in the response. If a validator flagged headline_quantified, no_throat_clearing, or similar, FIX YOUR PROPS AND CALL persist_block AGAIN. Don't ship past validators — they catch the patterns operators have explicitly told us look bad.

Step 5 — Verify and report

After all blocks land:

a. Call get_workspace_snapshot(workspace_id) to confirm the landing page rendered (landing_pages.sections now has hero / services / etc instead of the original chatbot-preview).

b. Report the public URL back to the operator with a brief summary of what you placed and which archetype voice you used.

c. Offer concrete next steps:

• "Want a different hero photo? Tell me what to search for."
• "Want a softer voice? I can rerun in editorial-warm."
• "Want to add a testimonials block? Share the testimonials."
• "Want to publish the chatbot to LIVE now that the page exists? Just say so."

Anti-patterns — DO NOT DO

• Don't skip get_workspace_state. You need the archetype, soul, and theme. Guessing wastes round-trips when the server overrides your picks anyway.
• Don't write throat-clearing copy. "Welcome to" / "Your trusted" / "Professional X services" are banned per every block's validator.
• Don't propose templates outside the registry. Templates are cinematic-aura | viktor-light | velorah-editorial | nexora-light | securify-bold | stellar-tabs-white. Picking anything else gets overridden server-side. Bold-urgency archetypes intentionally use empty template (the legacy variant renderer).
• Don't generate Unsplash queries longer than 4 words. Long queries zero-result. The server has archetype-curated fallbacks but those defeat your intent. Stick to 2-4 word queries.
• Don't add features the workspace can't support. If integrations.twilio.configured is false, don't promise SMS in copy. If Stripe isn't connected, don't say "pay deposit online."
• Don't reorder blocks across persist calls. Each persist_block call REPLACES that block type — order is determined at first persist. If you change your mind on order, call update_landing_content with the full new sequence.

Worked example — bold-urgency (HVAC plumbing)

Operator: "build a landing page for ignitify in bold-urgency style"

1. get_workspace_state(ws-1) → archetype: bold-urgency, vertical: hvac, business_name: Ignitify Cooling, services: [AC Repair, AC Install, Furnace Repair, Maintenance], reviews: 13, emergency_service: true

2. Block sequence: hero → emergencyStrip → servicesGrid → faq → cta (skip testimonials since reviews < 50; add emergencyStrip per bold-urgency rule; add stickyMobileCTA)

3. get_block_skill("hero") → load voice rules (Hormozi-style, quantified, urgent), prop schema (headline / subhead / ctaPrimary / background_image_query / variant / template)

4. Generate hero props:

   {
     "headline": "Same-Day AC & Furnace Repair Across El Paso",
     "subhead": "Ignitify Cooling — BBB-accredited, SuperPros 2024 Gold technicians who fix it right the first time. Honest pricing, financing available.",
     "ctaPrimary": { "label": "Get Service Today", "href": "/book" },
     "ctaSecondary": { "label": "Free Estimate", "href": "/intake" },
     "background_image_query": "hvac technician outdoor",
     "variant": "split-screen-50-50"
   }
   

   (Server will override variant to "split-screen-50-50" anyway since archetype is bold-urgency — your pick happens to match.)

5. persist_block(ws-1, "hero", "...", props) → returns ok + no warnings → continue to emergencyStrip

6. Repeat steps 3-5 for each subsequent block.

7. get_workspace_snapshot(ws-1) → confirm 5 sections rendered.

8. Report:

   Landing page is live at https://ignitify-cooling.app.seldonframe.com. Rendered in bold-urgency voice with 5 sections: hero, emergency strip, services grid, FAQ, closing CTA. Hero uses split-screen-50-50 layout with service-truck imagery. Want to tweak anything?

Worked example — clinical-trust (dental practice)

Skip emergencyStrip. Lead with credentials in the hero subhead. Use nexora-light template (archetype default). Lengthen FAQ with insurance

• financing + "do you accept Medicare" type questions.

[Full prop JSON for each block omitted for brevity — follow the same pattern as the bold-urgency example with the clinical-trust voice: calm, authoritative, precise.]

Worked example — cinematic-aspirational (medspa)

Lead with projectGallery immediately after hero. Hero uses cinematic-aura template (archetype default) with a Pexels video background. Soften FAQ to "what to expect" / "is it painful" tone.

[Full prop JSON for each block omitted for brevity — follow the same pattern with the cinematic-aspirational voice: sensory, restorative, intentional.]

Integration notes

• v1.54 archetype enforcement still fires. Don't worry about picking the exact right template — the server overrides if you're off. Just match your COPY to the archetype voice.
• Brain v2: before generating, optionally call list_brain_patterns(workspace_id) to see what's worked for similar verticals. If brain patterns exist for "vertical=plumbing" (e.g., "service-truck hero photos performed best"), fold them into your block generation.
• Validator gates are real. If persist_block returns warnings, fix props and call again. Don't ship past validators.