Performance

Turn a speed test report into a prioritized fix list

A scenario prompt that diagnoses a real 41/100 PageSpeed report against a specific shared-hosting, page-builder stack, ranking fixes by what's actually feasible, not just what the report lists first.

Works with Claude / GPT720 uses 4.5

The prompt

pagespeed-report-to-fix-list
A client's WooCommerce store just scored 41/100 mobile on PageSpeed Insights. Here's
what the Diagnostics section flagged:

[- Eliminate render-blocking resources: 1.2s savings, mostly from a slider plugin's
CSS loading in <head> on every page including ones without a slider
- Reduce unused JavaScript: 800ms, largely from a page builder's frontend bundle
loading sitewide
- Largest Contentful Paint image not preloaded: hero image on homepage]

HOSTING: [shared hosting, no server-level caching, PHP 8.1]
ACTIVE PLUGINS: [WooCommerce, Elementor, a slider plugin, a caching plugin set to
default settings]

Work through this as a performance consultant would, not by relisting the report:
1. Rank the three findings by real-world impact versus how risky the fix is to
   implement, given this is shared hosting with no dev-level server access. That
   changes which fixes are even available compared to a VPS.
2. For each, name the specific likely cause given the plugin stack. The slider CSS
   loading sitewide, for example, is almost certainly the slider plugin not
   conditionally enqueuing its assets, so name the setting or code fix, not just
   "check the slider plugin."
3. Give the exact fix: a plugin setting, a snippet using `wp_dequeue_style` or
   `wp_dequeue_script` scoped correctly so it doesn't break the slider where it's
   actually used, or a caching plugin setting to enable.
4. Flag which of these, if any, actually need a hosting upgrade (if the caching
   plugin is already maxed out and TTFB is still slow, for example) versus which are
   fixable purely from wp-admin.
5. Warn specifically about the risk of over-aggressively dequeuing the page builder's
   JS sitewide. Elementor-built pages elsewhere on the site could break if the fix
   isn't scoped per-template.

Two sites with the identical “eliminate render-blocking resources” warning can need completely different fixes depending on hosting and plugin stack. That’s why this reasons from the stack, not the raw score, and why step 5 exists: the most common way these fixes go wrong is a global JS deferral that quietly breaks a builder-made page somewhere else on the site.

Pro tip: paste the full “Diagnostics” section from PageSpeed Insights, not just the score. The score alone gives this nothing to reason about.