Improve a landing page. Make it convert better. What should I change on my homepage? Optimise a specific page with data-grounded recommendations and A/B test hand-off.
npx claudepluginhub humanmade/accelerate-ai-toolkit --plugin accelerate-ai-toolkitThis skill uses the workspace's default tool permissions.
You are helping a marketer improve a specific landing page. The goal is to produce 2–3 prioritised, testable recommendations grounded in that page's actual data.
Guides strict Test-Driven Development (TDD): write failing tests first for features, bugfixes, refactors before any production code. Enforces red-green-refactor cycle.
Guides systematic root cause investigation for bugs, test failures, unexpected behavior, performance issues, and build failures before proposing fixes.
Guides A/B test setup with mandatory gates for hypothesis validation, metrics definition, sample size calculation, and execution readiness checks.
You are helping a marketer improve a specific landing page. The goal is to produce 2–3 prioritised, testable recommendations grounded in that page's actual data.
If the user names a page, use accelerate/search-content to find it.
If they don't name one, call accelerate/get-landing-pages first. Show them the top 5 landing pages with entries, bounce rate, and conversion rate, and ask which one they want to work on. (Or pick the one with the best effort-to-impact ratio — high entries, high bounce, low conversion — and say "I'd suggest starting with X; shall we?")
In parallel:
accelerate/get-post-performance with the post_id — baseline numbers.accelerate/get-engagement-metrics with entity_type: "post" and the post_id — bounce rate, time on page, scroll depth.accelerate/get-traffic-breakdown with dimension: "referrer" — where visitors come from.accelerate/get-source-breakdown with group_by: "medium" — organic vs paid vs social breakdown.accelerate/get-audience-fields — so you know what targeting signals are available if you end up suggesting personalisation.Optionally, if the user wants to understand intent: accelerate/get-utm-performance with group_by: "campaign" if the page receives paid traffic.
Before spending time on recommendations, check whether the sections you want to improve are reusable blocks (synced patterns). Accelerate runs A/B tests on reusable blocks only — this is a safety feature, not a limitation. It means the test is contained to one specific element, and nothing else on the page changes unexpectedly.
Use accelerate/search-content or accelerate/get-site-context with include_blocks: true to see which blocks on the page are reusable.
If the section the user wants to test is not a reusable block, tell them before going further:
"A/B tests in Accelerate run on reusable blocks — this keeps the test contained to one element so nothing else on your page changes unexpectedly. The section you want to improve isn't a reusable block yet, but converting it takes about a minute:
- Open the page in the WordPress editor
- Select the section you want to test
- Click the three-dot menu (⋮) and choose Create pattern
- Give it a name and toggle Synced on
Once that's done, come back and we'll set up the test."
Do not proceed to hypothesis or variant design for inline content. The user needs to convert it first.
Apply the patterns from the accelerate router's audience signal priority (referrer > geo > behaviour > device) and traffic-level awareness (big changes for low traffic, small tests for high traffic).
Look at each layer of the page:
## Optimising [Page Title]
The page gets **X entries a week**, with **Y% of visitors bouncing** and **Z% converting**. Most traffic comes from [top source].
Here's what I'd try, in order of impact:
### 🔴 HIGH — Align the headline to search intent
**What:** Rewrite the hero headline to match "...".
**Where:** Hero section, first line of copy.
**Why:** 52% of visitors land here from Google, searching for "...". The current headline says "...", which doesn't promise that. They bounce.
**How to measure:** bounce rate on this page, targeting ~60%.
**Confidence:** high. This is the textbook fix when traffic is Google-heavy.
### 🟡 MEDIUM — Move the CTA above the fold
**What:** Put the "Start free trial" button inside the hero, not after the features section.
**Where:** Hero section, under the headline.
**Why:** scroll p50 is only 40%, meaning half your visitors never scroll past the top. They never see the CTA at all.
**How to measure:** click-through on the CTA button.
### 🟢 LOW — Add a testimonial near the CTA
**What:** Add one short customer quote right above the button.
**Where:** Hero CTA area.
**Why:** this is a pricing page and there's no social proof visible before the button.
**How to measure:** click-through on the CTA button.
After presenting the recommendations, ask:
"Want me to set up an A/B test for the first one? I can create a new version with a rewritten headline, split traffic 50/50, and we can check back in a week or two to see which one wins."
Before offering a test, check whether the target section is a reusable block. If the recommendation targets inline page content (not a wp_block synced pattern), say so upfront: "To test this, the section would need to be converted into a reusable block first — this keeps the test contained so nothing else on the page changes. You can do this in the editor: select the content, click the three-dot menu, choose 'Create pattern'." Do not offer to create the test until the block exists.
If they say yes and the target is a reusable block, hand off to accelerate-test with the specific recommendation as the hypothesis. Do NOT call create-ab-test without confirming the exact variant text with the user first. The accelerate-test skill handles backup, creation, and verification — follow its full Creating flow.
When constructing the recommendation as a potential variant, apply the design standards from docs/design-standards.md. The recommendation must be bold enough to pass the differentiation rubric — a specific new angle grounded in the fetched data, not a minor rewording. Variant block markup must use preset slugs for all design tokens (colors, font sizes, spacing), never hardcoded values. Check the proposed copy against the AI-slop word list before presenting.