Landing Page Performance Audit
Purpose
Run a standalone, evidence-based audit of the Performance module for an existing landing page. The purpose is to diagnose the current experience and recommend practical improvements, not to build a replacement page.
How to run this skill independently
This skill audits only its named module. Do not assume any other module has been run. Begin with the user's request and supplied materials. If essential context is missing, ask only for information that materially affects this module; do not block on analytics or details that are unavailable. If the page URL is inaccessible, request screenshots, a recording, exported page text, or HTML. Work within the evidence available and label the scope limitation.
Useful context, when available: business/brand and industry; target audience and geography; page URL or assets; product/service and business model; primary conversion goal; offer/price; campaign or traffic source; brand constraints; relevant competitors; and analytics/testing data. Mark missing data Not provided or Unavailable rather than guessing. Distinguish business-provided claims from your own page observations.
Evidence and integrity rules
- Separate Observed (directly visible/tested), Reported (provided by the business), Inference/Hypothesis (plausible but unverified), and Recommendation.
- Tie important findings to a page section, element, source URL, supplied asset or test result. Quote exact copy only if it is available; otherwise label a paraphrase.
- Explain what is present, why it matters to the relevant user/task, and what to change. Avoid treating every weakness as a proven conversion loss.
- Do not invent business facts, results, analytics, guarantees, user research, certifications, or test outcomes. Avoid causal and uplift claims unless supported by valid experiments.
- If only static screenshots or text are available, do not claim to have tested links, forms, responsive behavior, performance, keyboard access, or assistive technology.
- Consider audience, offer, conversion goal, traffic context and accessibility needs where supplied. State assumptions and material limitations.
Priorities, confidence and scoring
Use priority P0 Critical (serious broken/usability/trust issue requiring immediate attention), P1 High (important, potentially consequential), P2 Medium (valuable after higher priorities), P3 Low (minor refinement/longer-term). Use confidence High (directly observable/measured), Medium (reasonable evidence with gaps), or Low (limited evidence or substantial assumptions).
Score applicable criteria from 0–5: 0 missing/broken/critically deficient; 1 very poor; 2 below expectations; 3 acceptable; 4 very good; 5 excellent. Mark unassessable items N/A and exclude them from the denominator. Calculate the module score as (points earned ÷ maximum applicable points) × 100, rounded to a whole number; state the numerator/denominator or list scored criteria. If nothing can reasonably be scored, report Not Evaluated, not 0. Do not treat this diagnostic score as an industry benchmark or a forecast of conversion performance.
Output format
Write a self-contained, client-ready Markdown audit with a title, audit scope/evidence basis, concise summary, strengths, prioritized findings/recommendations (prefer a table where useful), module score, overall evidence confidence, and limitations/next evidence needed. Keep advice specific and actionable. Do not produce a full multi-module report or score unless the user explicitly asks for one.
Module-specific review checklist
Review only available evidence about load timing, Core Web Vitals, image/font delivery, JavaScript/CSS, render-blocking resources, layout shifts, third-party scripts, lazy loading and network/resource efficiency. Record URL, date/time, device/network, tool, test conditions, and whether results are lab or field data when known.
Required deliverable
Report measurements with units and provenance; distinguish measured findings, visual observations and hypotheses; explain user impact cautiously; recommend specific next diagnostics/fixes; assign priority/confidence and score. If no technical measurements are available, mark performance findings unverified and explain how to obtain evidence.
Module-specific guidance
Reference thresholds: LCP good ≤2.5 s, needs improvement >2.5–4 s, poor >4 s; INP good ≤200 ms, needs improvement >200–500 ms, poor >500 ms; CLS good ≤0.1, needs improvement >0.1–0.25, poor >0.25. Use exact boundary consistently. Do not invent metrics or claim a pass without valid measurements. A screenshot cannot establish speed.
Strategic diagnostic questions
- Which performance conclusions are backed by field/lab measurements, and under what device, network, region and test conditions?
- Is the proposed fix connected to an observed bottleneck, or is it a generic optimization checklist?
- Could a third-party script, oversized asset, font, or layout shift affect a critical action; what evidence would isolate it?
- What should be measured after remediation, and how will lab results be distinguished from real-user outcomes?
Final quality check
Before delivering, confirm that the review stayed within this module, every material conclusion is evidence-labelled, all recommendations are actionable and appropriately prioritized, unknowns remain unknown, N/A items were excluded from scoring, and the score calculation is correct. If the user asks for a full 12-module audit, explain that this skill covers one module and can be run alongside the other skills in this repository.