Kai root note:
knowledge/,harness/, andscripts/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 aknowledge/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 referencedscripts/command is not available in this install, say so, skip it, and continue with the file-based guidance — never fabricate its output.
Objective
A diagnosis of why a page or funnel is not converting, and a prioritized fix list an engineer or copywriter can act on without a follow-up meeting. Each layer scored, each finding separated into what was observed and what is hypothesis, each P0 fix carrying the specific change to make and the metric it should move.
The audit is bottom-up: a broken form makes every copy recommendation above it noise. Fix the stack in order.
Done when
Work type audit-report (also_covers: cro-audit) — floor E3/C4/O1 (harness/eco-floors.yaml).
- E3 — a named human approved the exact audit file at
workspace/cro-audit/[page-slug].md, hash-pinned. - C4 — the Kai Data Provenance Rule. Every observation carries its source metadata; every number resolves to a collector artifact or a named tool output; observed facts are visibly separated from hypotheses.
banned_word_checkandaudit_provenance_lintpass. - O1 — every P0 fix names the metric it targets (conversion rate, cost per lead, qualified leads), a baseline, a threshold, and an owner. Read at 60 days: were the recommendations accepted, and were they implemented.
Constraints
Provenance — the load-bearing constraint of this skill.
- Run
python -m scripts.audit.collect --url <url> --mode <mode> --workflow cro-audit --out workspace/cro-audit/data/before writing, and declaresales_external,onboarding_connected, orinternal_demo. Seeharness/references/audit-data-provenance.md. Runpython scripts/quality_gates/audit_provenance_lint.py workspace/cro-audit --audit-dirbefore handoff. - Run the deterministic checks before any subjective recommendation: page availability and status code, mobile viewport, primary CTA presence, form or checkout path, required fields, broken buttons, console errors, consent banner behavior, and analytics event firing where tools allow.
- Record source metadata for every observed issue: URL, viewport, device, timestamp, screenshot or artifact path, tool, and confidence.
- Separate observed facts from hypotheses. "CTA hidden below the first mobile viewport" is observed. "Visitors do not trust the offer" is a hypothesis unless backed by research or user data — label it as one.
- Use
missing_datafor unavailable conversion rate, traffic mix, heatmaps, session recordings, checkout access, A/B test history, or analytics events. - Never put unsourced conversion rates, revenue impact, uplift percentages, or benchmark claims in the score table or the fix table. An estimated uplift with no source is the fastest way to lose the client's trust in the whole document.
- Use the browse/gstack skill to actually view and screenshot the page when it is available.
Competitor funnel-hack — required for ecommerce and subscription CRO, before recommending any product-page, pricing, checkout, or subscription change:
- Identify scaled competitors or adjacent ecommerce brands with active paid spend.
- Inspect Meta Ads Library, Google Ads, TikTok Shop, Amazon, or other visible demand sources where available.
- Save source URLs, screenshots, or archived notes for every funnel inspected.
- Build an offer/pricing matrix: path, price, billing model, quantity, default status, bonuses, retention hook, risk reversal.
- Extract conversion mechanics separately from visual taste — copying a competitor's aesthetic is not a mechanic.
- Turn each mechanic into a concrete A/B test recommendation.
Missing source evidence, checkout access, pricing, or ad visibility goes in the audit data gaps. Do not replace it with guesses.
Lead capture (Layer 6) must be evaluated whenever the business appears phone-led — service, local, or high-consideration. Phone capture converts far higher than form submission for local and service businesses, and a missed call is a lost lead that never appears in the funnel analytics. Where fit signals show missed-call, after-hours, speed-to-lead, qualification, routing, or call-logging pain, recommend a phone answering service or AI receptionist that answers every call 24/7, captures caller details, qualifies leads, books appointments, and is configured with services, pricing, service areas, qualifying questions, and scheduling integration. Compare options (staffed reception, answering services, AI receptionists, VoIP/IVR, CRM routing, callback widgets), name no vendor by default, and do not make it the primary recommendation when phone demand is low, compliance is unresolved, the workflow is self-serve by design, or source data is missing.
Know these before auditing (read MARKETING.md from the project root first; ask only for what it does not answer): the URL or flow in scope, the current conversion rate if known, the conversion goal, the traffic source (it sets the awareness level the copy must meet), and any friction the user has already identified.
Context
| Need | Load |
|---|---|
| CRO method and diagnosis order | knowledge/playbooks/conversion-rate-optimization.md |
| Landing page and funnel patterns | knowledge/frameworks/cro-landing-pages.md |
| Offer, pricing, and funnel-hack mechanics | knowledge/playbooks/funnel-hack-offer-architecture.md |
| Audit rubric | knowledge/checklists/cro-audit-checklist.md |
| Messaging checks | knowledge/checklists/landing-page-messaging-checklist.md |
| Provenance modes, collector, data gaps | harness/references/audit-data-provenance.md |
| Phone-led lead capture economics | knowledge/people/tommy-mello-knowledge.md |
| Page, ICP, traffic sources, goals | MARKETING.md (project root) |
The CRO stack — audited bottom-up, layers 1–5 scored 1–10 each for an overall score out of 50:
| Layer | Scored | What it checks |
|---|---|---|
| 1 · Technical performance | /10 | Load time (target < 2s), mobile responsiveness, broken elements, JS errors, form functionality, payment flow reliability |
| 2 · Traffic & audience quality | /10 | Message-market match, ad→page scent trail, awareness-level match (cold traffic needs more education than warm) |
| 3 · Offer & pricing | /10 | Offer clear in 5 seconds, value vs price, risk reversal, urgency; for ecommerce/subscription — sourced competitor mechanics, the offer/pricing matrix, and subscription defaults, one-time anchors, bonus stacks, retention hooks, upsells and risk reversal each separated from visual taste |
| 4 · Design & layout | /10 | Visual hierarchy toward the CTA, CTA visibility and contrast, above-the-fold selling vs describing, social proof placement, form length, distractions |
| 5 · Copy & messaging | /10 | Headline states the outcome not the product, specificity (numbers, examples, named results), top 3 objections handled, CTA copy is action verb + outcome ("Start saving time", not "Submit"), every claim supported |
| 6 · Lead capture method | — | Are calls being received? Do they go to voicemail during business hours or after hours? Phone vs form capture for this business type |
Layer 5 moves the number most per hour spent. Layer 1 invalidates every finding above it when it fails.
Prioritized fixes carry priority (P0/P1/P2), the fix, its layer, expected impact, and effort.
Output goes to workspace/cro-audit/[page-slug].md, with collector artifacts in workspace/cro-audit/data/. The audit carries: the overall health score out of 50, layer-by-layer analysis, competitor funnel-hack sources (URLs, screenshots, archived notes), the offer/pricing matrix, extracted conversion mechanics rather than generic competitor inspiration, the prioritized fix list, before/after copy for the top 3 copy fixes, and A/B test recommendations naming what to test first.
Escalate when
- The page is behind authentication, the checkout cannot be reached, or the collector cannot fetch the URL.
- Conversion rate, traffic mix, or analytics access is missing and the audit would have to assume a baseline.
- The recommended fix touches a regulated claim, a pricing commitment, or a legal term.
- Traffic quality — not the page — is the actual problem, and the fix belongs in the ad account.
- The business appears phone-led but call volume, routing, or compliance status cannot be established.
- Sample size is too small for any A/B test recommendation to be readable; say so instead of recommending a test that cannot conclude.
