Communitygithub.com

local-landing-page-review-local-ads

Reviews a landing page for a local business, from page text and phone screenshots, checking the headline, area served, phone and booking path on the first screen, proof, offer, loading speed and tap targets, and ranking fixes by effort. Use when someone types "review my landing page", "local landing page review", "does my page convert local customers", or "what is wrong with my service page". Not for a form-only review; use contact-form-friction-review-local-ads.

Qu'est-ce que local-landing-page-review-local-ads ?

local-landing-page-review-local-ads is a Claude Code agent skill that reviews a landing page for a local business, from page text and phone screenshots, checking the headline, area served, phone and booking path on the first screen, proof, offer, loading speed and tap targets, and ranking fixes by effort. Use when someone types "review my landing page", "local landing page review", "does my page convert local customers", or "what is wrong with my service page". Not for a form-only review; use contact-form-friction-review-local-ads.

Compatible avec~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/mardab96/local-business-ads-claude-skills/tree/HEAD/local-landing-page-review-local-ads

Demander à votre IA préférée

Ouvre une nouvelle conversation avec cette compétence d'agent déjà préchargée.

Documentation

Local Landing Page Review

Use this skill when

A local business sends ad traffic to a service page or a home page, and leads or calls are weak. This skill reads the page text and a phone screenshot and checks what a local visitor needs: what you do, where, how to reach you, and why to trust you. It ranks fixes by effort. It never edits the page.

Common requests:

  • "Review my landing page."
  • "Local landing page review."
  • "Does my page convert local customers?"
  • "What is wrong with my service page?"

Send these jobs to another skill:

  • Ad promise versus page: ad-to-page-promise-match-local-ads.
  • Form: contact-form-friction-review-local-ads.
  • Phone button: mobile-click-to-call-check-local-ads.
  • Form event tracking: form-lead-event-audit-local-ads.

Required input

Work from pasted exports and screenshots. No live account access is needed. If an input is missing, stop that finding, name the missing input, return needs_data, and do not estimate.

Required

  • Page URL and the page text from top to bottom.
  • A phone screenshot of the first screen without scrolling.
  • Main action the page should drive: call, form, or booking.

Optional

  • A PageSpeed or Search Console export with the Largest Contentful Paint value (T28), either field data or a lab run.
  • Reviews, credentials, licence numbers, and guarantees the business holds.
  • Service area as the business defines it.

Export route:

  1. Open the page on a phone. Screenshot the first screen, then copy the page text.
  2. Run PageSpeed Insights on the URL and note the Largest Contentful Paint. Note whether it is field data or a lab run.
  3. List the proof the business can show.

See ../local-ads-shared/references/getting-your-data.md.

Analysis workflow

Read ../local-ads-shared/references/thresholds.md before step 1.

  1. State the page, the ad source it serves, and the main action.
  2. Read the first screen on a phone. T30 asks for a tappable phone number and the area served (or an address) there. Also check that the service and the main action are visible without scrolling. Mark each confirmed, partial, or missing.
  3. Check the phone. It must be a tappable link on the first mobile screen. Route detail to mobile-click-to-call-check-local-ads.
  4. Check proof: reviews with a source, licence or registration, insurance, years in business, and photos of real work. Proof the page cannot show is not invented.
  5. Check the offer: price or price range, estimate terms, response time, and guarantee. A visitor should not have to ask what happens next.
  6. Check speed. Largest Contentful Paint should be at or below the T28 value at the 75th percentile. Field data gives a likely verdict. A PageSpeed lab run gives a possible verdict with the evidence tag hypothesis (lab data). With no figure at all, return needs_data for speed.
  7. Check tap targets on the screenshot (T29). A small link or a crowded button row costs taps.
  8. Check the form separately. A long form is a finding for contact-form-friction-review-local-ads.
  9. List fixes ranked by effort and impact: text edits first, then layout, then new content. Each fix is a proposal for the page owner.
  10. Fill the decision table. Findings drawn from text or a screenshot are confirmed. A guess about why visitors leave stays hypothesis.

Decision rules

Thresholds live in ../local-ads-shared/references/thresholds.md.

CheckPass conditionIf failedIf it cannot be evaluated
Service and area on the first screen (T30, plus the service)Area served and service both visible without scrollingconfirmed findingneeds_data
Phone on the first screen (T30)Tappable number visibleconfirmed findingneeds_data
ProofAt least one verifiable item near the main actionlikely low trustneeds_data
OfferEstimate terms and next step visiblelikely frictionneeds_data
Speed (T28)Largest Contentful Paint at or under the thresholdlikely slow page from field data, possible (hypothesis (lab data)) from a lab runneeds_data when no figure is given
Tap targets (T29)Targets meet the platform guidepossible tap errorsneeds_data
PatternVerdictNext step
Home page used for a service adlikely weak matchad-to-page-promise-match-local-ads
Phone only in the footerconfirmedmobile-click-to-call-check-local-ads
Long form is the only pathlikelycontact-form-friction-review-local-ads
Slow page on mobilelikelyPropose speed fixes for the developer
Page fine, leads weakpossible traffic problemsearch-term-intent-filter-local-ads

A page review does not prove a page converts. A change is judged only after volume (T01) and the booking lag (T02).

Output format

Use labels and tags from ../local-ads-shared/references/output-standard.md.

## Local landing page review: [account], [URL]
Main action: [call / form / booking] | Traffic from: [ad source]

### Checks
| Check | Evidence | Result |
|---|---|---|

### Decision table
| Finding | Evidence | Verdict | Severity | Confidence | Business impact | Next step |
|---|---|---|---|---|---|---|

### Fixes ranked by effort
| Fix | Effort | Impact | Owner |
|---|---|---|---|

### Data gaps
[missing screenshots or exports and the verdict each would firm up]

Practical example

Illustrative only: made-up page for an HVAC company.

CheckEvidenceFindingVerdict
Area and service (T30)Headline says "Quality comfort for your home". No area named on the first screenArea missing, service not namedconfirmed
Phone (T30)Phone number is an image in the footerNot tappableconfirmed
ProofReviews are three unsourced quotes. A licence number is in the footerLow trustlikely
Speed (T28)PageSpeed lab run only, Largest Contentful Paint over the thresholdSlow in a lab runpossible, evidence tag hypothesis (lab data)

Recommendation: change the headline to the service and the area, add a tappable number to the first screen, and move the licence number near the call to action. All three are text edits, so low effort. Ask for field data before the speed finding goes above possible. Each is approval_needed for the page owner.

Common failure modes

  • Reviewing the design without reading the ad.
  • Inventing proof the business does not hold.
  • Giving a speed verdict above possible without field data.
  • Treating a page review as a conversion result.
  • Rewriting the whole page instead of the first screen.

Route to next

Ad match: ad-to-page-promise-match-local-ads. Form: contact-form-friction-review-local-ads. Phone: mobile-click-to-call-check-local-ads. Tracking: form-lead-event-audit-local-ads.

Guardrails

  • Never change a live account, budget, bid, tag, page, price, or form.
  • Never invent numbers. Missing data becomes needs_data.
  • End in an approval-ready recommendation for a named human decision.
  • Never edit a page or publish copy.
  • Never invent reviews, credentials, prices, or guarantees.
  • Never give a speed verdict above possible without field data. A lab run is hypothesis (lab data).
  • Never request or echo customer names, emails, or phone numbers. Ask for them removed before per-customer rows are pasted.
  • If the account is ecommerce or SaaS rather than local lead generation, say so and stop.

Skills associés