page-speed-impact-review
Skill2 starsChanged 4 months agoMITView on GitHub
Turns speed measurements into a conversion priority instead of a scoreboard, and says which specific asset to fix first. Use when a speed report has been run and nobody knows whether any of it is worth the engineering time.
In this repository26 other files it gives its agents.
- AGENTS.mdAGENTS.md
- above-the-fold-clarity-reviewSkill
- accessibility-conversion-blocker-checkSkill
- comparison-page-positioning-reviewSkill
- conversion-leak-finderSkill
- cta-clarity-checkSkill
- form-friction-finderSkill
- google-ads-landing-page-experience-reviewSkill
- hero-section-diagnosisSkill
- landing-page-ab-test-readoutSkill
- landing-page-copy-readability-passSkill
- landing-page-scale-readiness-checkSkill
- landing-page-triageSkill
- lead-form-sales-handoff-checkSkill
- mobile-conversion-reviewSkill
- objection-map-builderSkill
- offer-clarity-diagnosisSkill
- page-length-fit-checkSkill
- paid-traffic-message-match-auditSkill
- popup-and-overlay-timing-reviewSkill
- pricing-page-clarity-reviewSkill
- social-proof-strength-auditSkill
- thank-you-page-opportunity-auditSkill
- traffic-temperature-match-reviewSkill
- trial-vs-demo-path-decisionSkill
- trust-signal-auditSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
No report and no way to run one: stop. This is the one skill in the pack that has nothing to say without a measurement, and saying so is …
- Very specific: line 32
Not for: mobile layout and interaction friction (mobile-conversion-review), or the Google Ads landing page experience rating …
- Distinctive: line 48
Never state a conversion lift from a speed change.
- Distinctive: line 49
Do not recommend a rebuild, a framework change or a CDN purchase when the report points at three assets.
- Distinctive: line 63
No predicted conversion lift is given, and say why
- Distinctive: line 75
Do not run without a measured report.
- Distinctive: line 76
Do not predict a conversion lift from a speed change.
- Distinctive: line 77
Do not recommend a rebuild or a purchase when named assets are the problem.
- Distinctive: line 78
Always say whether a number is lab or field.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Page Speed Impact Review
Use this skill when
- a measured speed report exists, from PageSpeed Insights, Lighthouse, Core Web Vitals field data or a real device test.
- a speed report has been run and nobody knows which line of it is worth engineering time.
- images or third-party tags were added and nobody measured what they cost.
This skill needs a measured report. Without one it names the tool to run and stops.
Required input
- a measured speed report: PageSpeed Insights, Lighthouse, Core Web Vitals field data, or a real device test. Name the tool and whether the numbers are lab or field.
- the specific offending assets if the report names them: image sizes, blocking scripts, fonts, third-party tags.
- device split and the conversion rate per device, so priority can follow the money.
- If no measured report exists, this skill does not run. Say which tool to run and stop. An impression of slowness is not an input.
If the data does not exist
Getting the report takes two minutes and there is no substitute for it.
- Field data, which is what matters: pagespeed.web.dev, enter the URL, and read the Core Web Vitals Assessment section at the top. That is real users at the 75th percentile. It only appears if the page has enough traffic; if it says there is not enough data, say so rather than substituting the lab score.
- Lab data, for diagnosis: the Lighthouse panel in Chrome DevTools, mobile preset, throttled. It names the LCP element and the largest assets, which is what turns a score into one fix.
- Asset sizes: the Network panel, sorted by size, with the cache disabled and the page reloaded.
- No report and no way to run one: stop. This is the one skill in the pack that has nothing to say without a measurement, and saying so is more useful than a plausible guess about a hero image.
Not for: mobile layout and interaction friction (mobile-conversion-review), or the Google Ads landing page experience rating (google-ads-landing-page-experience-review).
Analysis workflow
- Collect speed signals: Core Web Vitals, load time, mobile speed, script weight and observed drop-off.
- Separate perceived speed blockers from actual conversion blockers.
- Identify heavy scripts, media, third-party tags, forms or layout shifts likely to hurt paid traffic.
- Compare speed issue severity with device mix and landing page conversion data.
- Recommend speed fixes only where they plausibly affect conversion or tracking reliability.
Decision rules
- Field data outranks lab data for prioritisation, and lab data outranks field data for diagnosis. Say which one each number is; a Lighthouse score is not what users experienced.
- Identify the LCP element before recommending anything. One oversized hero image is a single fix; a general speed programme is a quarter of engineering time.
- Layout shift is a conversion defect, not a technical one. Content moving under a thumb produces mis-taps on the elements that convert, so report it in those terms.
- Priority follows traffic share and conversion value, not the size of the number. A 5.8 second LCP on the 12 percent of traffic that is desktop is a smaller problem than a 3 second one on mobile.
- Never state a conversion lift from a speed change. Attribution needs everything else held still, and it never is. Say hypothesis.
- Do not recommend a rebuild, a framework change or a CDN purchase when the report points at three assets.
Output format
| Metric | Value | Lab or field | Percentile | Offending asset | Fix | Effort |
| LCP / CLS / INP / TTFB / weight | As measured, with units | Lab / field | e.g. 75th | Named asset or unknown | The specific change | One asset / one template / engineering programme |
End with:
Highest-value single fix:the one change, and the metric it movesTraffic exposure:device share carrying the worst numbersDecision:fix the named assets / measure field data first / escalate to engineering / ask for the reportMissing data:what the report did not include- No predicted conversion lift is given, and say why
Practical example
User: "PageSpeed Insights on mobile: LCP 5.8s, field data, 75th percentile. CLS 0.24. Hero image is a 2.4MB uncompressed PNG. Three font files load before first paint. Mobile CVR 0.9 percent and mobile is 70 percent of traffic."
Assistant should note: LCP at 5.8s on the 75th percentile of field data means a quarter of real mobile visitors wait longer than that, and the 2.4MB hero PNG is almost certainly the LCP element, which makes this one fix rather than a programme of work; CLS at 0.24 means the layout moves under the visitor's thumb after paint, which produces mis-taps on exactly the elements that convert; 3 fonts blocking first paint compound the same problem.
Then: compressing and correctly sizing one image is the highest-value change available, and mobile carrying 70 percent of traffic is what makes it a business priority rather than a technical chore. What this refuses to give is a predicted conversion lift: attributing a CVR change to speed needs everything else held still, and it never is, so the effect on the 0.9 percent stays a hypothesis. Whether the PNG is genuinely the LCP element is unverified until the report names it; if it does not, that identification is the first thing to confirm.
Guardrails
- Do not run without a measured report.
- Do not predict a conversion lift from a speed change.
- Do not recommend a rebuild or a purchase when named assets are the problem.
- Always say whether a number is lab or field.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

