/kai-landing-page — A Page That Converts the Traffic It Actually Gets

Kai root note: knowledge/, harness/, and scripts/ paths in this skill live in the Kai install, not the user's project. Resolve them against the first ancestor directory of this SKILL.md that contains a knowledge/ folder (the Kai plugin root, ~/.claude/kai, or the kai-cmo-harness repo). MARKETING.md, memory/, and any output files live in the current project. If a referenced scripts/ command is not available in this install, say so, skip it, and continue with the file-based guidance — never fabricate its output.

Objective

Complete, section-by-section landing page copy built for one traffic source and one conversion action — hero through FAQ, every claim carrying its proof, one CTA destination throughout. The page applies perception engineering in sequence: destabilize the cached belief, shift the frame, then remove the consequences of acting.

Awareness level is the load-bearing judgment. Cold ad traffic and warm referral traffic need different first screens, and a page written for the wrong one fails no matter how good the copy is.

Done when

Work type landing-page — floor E5/C3/O4 (harness/eco-floors.yaml, contract harness/skill-contracts/landing-page.yaml).

• E5 — the live page returns 200, renders the approved copy, and its primary CTA submits successfully, verified by someone other than the writer. Approved copy in a file is E3, not E5.
• C3 — four_us_score ≥ 12/16 page-level, banned_word_check at zero, seo_lint clean, and a named non-producer reads the page end to end.
• O4 — conversion rate, qualified leads, or cost per lead clears a threshold declared before ship, read from the analytics connector at 30 days. Do not read the metric before the declared minimum sample; an underpowered read is not an observation.

Attribution is required. A page replacing an existing page needs a control or holdout — see knowledge/frameworks/marketing-science/experiment-rigor.md.

Constraints

• The wireframe is approved before any copy is written. Present sections in order and get sign-off. Adapt the structure to the product; not every page needs every section.
• The variant lab runs before full copy, and its winner is approved before production. Produce 5 hero angles (headline, subhead, CTA, traffic source, awareness level), 3 offer frames (direct outcome, risk removal, urgency), 3 proof plans (testimonials, metrics, demos, screenshots, founder story), and a kill list naming why each rejected angle died. Score every variant on the rubric below; only variants scoring 20/25+ reach copy production. Save to workspace/landing-pages/_variant-lab.md.
• Every claim needs proof — a stat, a testimonial, or a specific example. A section with no available proof gets cut, not padded.
• One CTA destination across the whole page. Wording may vary per section; the destination may not.
• Phone Capture Fit Rule. Evaluate phone-based lead capture whenever the business appears phone-led. For service businesses (legal, medical, home services, contractors), phone calls convert far better than form fills, and the primary CTA should be "Call Now — Free Consultation" with a large, clickable, tracked number that is answered 24/7 (staff, an answering service, or an AI receptionist), qualifies the caller, and books appointments. Compare alternatives, and do not lead with it when phone demand is low, the workflow is self-serve by design, or compliance is unresolved.
• Copy rules, all sections: no banned words · no AI slop ("In today's rapidly evolving…") · sentences under 20 words · bold the benefit, not the feature name · one CTA per section.
• Hero rules: headline 6-12 words stating the outcome, not the product · subhead 15-25 words qualifying the audience and expanding the promise · CTA is an action verb plus an outcome ("Start closing leads", not "Sign up") · above the fold holds headline, subhead, CTA, one visual, nothing else.
• Perception layering: perception layer in the first two sections, context layer through the middle, permission layer in the closing sections.
• Validate against both checklists before handoff: knowledge/checklists/landing-page-messaging-checklist.md and knowledge/checklists/perception-engineering-checklist.md.
• Quantitative claims — conversion lifts, customer counts, funded amounts, review counts — follow the Kai Data Provenance Rule: collector-sourced or absent. See harness/references/audit-data-provenance.md.
• Kai writes copy. Publishing the page is a human action.
• Read MARKETING.md from the project root before asking the user anything. If it does not exist, build it from the codebase — README, manifests, landing pages, route files, analytics and email config — and confirm the draft. Do not open with discovery questions the repo can answer.

Know these before writing (from MARKETING.md first; ask only for what it cannot answer): traffic source and the awareness level it implies (cold ads / warm content / hot referral) · the conversion action (signup, demo, purchase, waitlist, phone call) · rewrite or net-new · what proof actually exists (testimonials, case studies, metrics, logos) · whether the business takes phone calls.

Context

Need	Load
The three persuasion layers	knowledge/frameworks/content-copywriting/perception-engineering.md
Messaging validation	knowledge/checklists/landing-page-messaging-checklist.md
Messaging workflow	knowledge/playbooks/landing-page-messaging-workflow.md
Conversion mechanics, friction, form design	knowledge/playbooks/conversion-rate-optimization.md
Page structure and CRO patterns	knowledge/frameworks/cro-landing-pages.md
Perception check before handoff	knowledge/checklists/perception-engineering-checklist.md
Format contract, word counts, gate thresholds	harness/skill-contracts/landing-page.yaml
Persona hooks and language	knowledge/personas/_persona-index.md
Product, ICP, voice, proof assets	MARKETING.md (project root)

Page architecture — section, job, and which persuasion layer it carries:

Section	Purpose	Perception layer
Hero	Hook + promise + CTA	Perception — destabilize cached beliefs
Problem	Agitate the pain	Perception — re-index virtues as vices
Solution	Your approach, not features	Context — shift what feels allowed
How It Works	3-step simplification	Context — genre-shift complex→simple
Social Proof	Testimonials, logos, metrics	Permission — remove risk
Features/Benefits	Outcome-framed, what they get	Context — expand possibility
Objection Handler	Top 3 objections	Permission — remove consequences
Pricing/CTA	Final push + urgency	Permission — future pacing
FAQ	Catch remaining doubts	Permission — double binds

Variant scoring rubric (1-5 each, carry 20/25+):

Factor	Question
Audience fit	Does it match the persona and traffic source?
Clarity	Is the offer understandable above the fold?
Proof readiness	Can the page support the claim with real evidence?
Differentiation	Does it avoid generic category copy?
Conversion fit	Does the CTA match user intent?

Output goes to workspace/landing-pages/[product-slug].md, with the variant lab at workspace/landing-pages/_variant-lab.md. Include the meta block (persona, traffic source, goal, word count) and gate results with the copy.

Escalate when

• The proof needed for a claim does not exist, and cutting the claim guts the page's core promise.
• Traffic source or awareness level is unknown — the page cannot be written against an unspecified visitor.
• The product is in a regulated category where the outcome claim needs substantiation the business has not supplied.
• The page replaces a converting page and no control or holdout is possible.
• Phone-led signals are present but call handling, recording consent, or routing is unresolved.
• Gates fail twice for the same reason — surface the specific failures and log the diagnosis in memory/lessons.md.