Fixed-10 Web Uplift evidence report

Catalog test design: HINT (not mandatory): a screenshot or computed background under an emulated prefers-color-scheme: dark condition will reveal whether surfaces re-tint; the page CSS / a color-scheme declaration is corroborating evidence. The model chooses the method.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video, or an in-page probe of getAnimations()/computed animation under an emulated prefers-reduced-motion: reduce condition, can show whether motion stops. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot under emulated prefers-contrast: more / forced-colors, or an axe/contrast probe, can show whether controls and text survive. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video of a route/state change shows whether it animates; the page source / ::view-transition usage corroborates. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source/CSS inspection for animation-timeline: scroll()/view(); a long-task / scroll-handler probe can flag the main-thread anti-pattern. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for scroll-snap / overscroll-behavior / physics-based easing vs custom pointermove listeners. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video of scrolling, or CSS inspection for scroll-state container queries. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for anchor-name / position-anchor / position-try on overlays; a screenshot of an open overlay near a viewport edge can show drift. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for ::highlight / scroll-marker; a transition video can show whether attention is cued after navigation. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot on load, or a DOM probe for full-viewport overlays present before interaction. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: DOM/source inspection for popover / <dialog> / <details> vs custom overlay divs with manual dismiss handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot plus layout metrics can show the proportion of the viewport given to chrome vs content. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: layout metrics (scrollWidth vs innerWidth) and a screenshot at an emulated narrow mobile viewport reveal overflow. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for @container / container-type; a computed-style probe of the same component in a wide vs narrow container shows whether it adapts. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a focus probe (focus an element, read the computed outline) or an axe target-size check; a screenshot of a focused control corroborates. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: screenshot the first viewport and key scrolled states; inspect heading structure, nav labels, button text, and visual hierarchy; a task walkthrough can show whether the next action is obvious. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: run the flow manually with screenshots/DOM snapshots at each step; compare expected vs actual path length; inspect form requirements, navigation continuity, and blockers. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise network delay/failure, invalid input, empty data and success states; screenshot the state messaging and recovery controls; inspect whether browser history and focus remain sensible. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse reports LCP/CLS/TBT directly and the model may run it; layout metrics + a layout-shift observer + a long-task observer (the evidence primitives) give the same signal first-party. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the layout primitive's CLS observer captures shifts; a transition video of the first seconds shows content jumping. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the layout primitive records long tasks; a heap summary shows the object population; Lighthouse reports TBT. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a HAR summary can reveal cache headers, redirects, render-blocking candidates, weight offenders and dependency shape; a trace/Lighthouse insight report can corroborate LCP discovery, render-blocking, font-display and document latency. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse/trace/code-coverage style evidence can flag unused JS/CSS, duplicated JavaScript and legacy code; a HAR summary shows third-party byte cost and request count. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe-core (injectable via the evaluate primitive) or Lighthouse's a11y audits enumerate these; a DOM probe of the accessibility-relevant attributes is a first-party alternative. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe contrast rules, a Lighthouse contrast audit, or an in-page probe computing contrast ratios from computed colours. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe/Lighthouse structural audits; a focus-walk probe (tab through, read activeElement + computed outline) is a first-party alternative. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot of body and heading text, plus CSS inspection for text-wrap / text alignment / font fallback handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse/axe target-size, meta-viewport and media-caption audits are useful signals; screenshots at narrow and zoomed conditions plus DOM/media inspection can corroborate. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: capture Runtime/Log CDP events, or a probe that reads collected errors; Lighthouse reports this too. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM/source probe for doctype/charset/img dimensions; CSS inspection for repetition; Lighthouse best-practices audits cover the rest. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse best-practices audits and DevTools inspector/deprecation signals can surface these; DOM/source probes can verify paste handlers and prompt timing. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe reads <title> and meta[name=description]; Lighthouse SEO audits cover the same ground. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe for anchor hrefs, viewport meta, and robots; Lighthouse SEO audits corroborate. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response status and headers, <link rel=canonical>, hreflang links, robots meta, robots.txt and sitemap.xml; Lighthouse SEO audits cover several of these. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect JSON-LD/microdata and social preview tags against visible content; Lighthouse has a manual structured-data audit, and ad-hoc probes can parse schema.org blocks. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response headers / page protocol via an evaluate probe or the network layer; Lighthouse best-practices flags HTTPS and CSP issues. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect network requests and third-party origins; a probe of analytics/beacon calls. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a probe for permission requests fired on load; source inspection for passkey / WebAuthn / navigator.credentials usage in auth flows. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response headers and browser security state; Lighthouse/DevTools security audits can corroborate HSTS, clickjacking, Trusted Types, origin isolation and third-party cookie findings. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: load with scripting disabled or compare a no-JS fetch of the HTML against the rendered page; check for Baseline-aware fallbacks in source. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise menus near viewport edges with a screenshot; a probe of async/visibility behaviour. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a probe for a service worker registration and a web app manifest; test behaviour offline. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: simulate failed fetches/offline mode or inspect representative 404/500 routes; screenshots and DOM snapshots of error/loading/empty states show whether recovery is possible. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe for <html lang>/dir and CSS inspection for logical vs physical properties. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for Intl.* usage vs hand-rolled formatting; a probe of rendered dates/numbers under a different locale. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for time-zone-aware date handling vs naive local Date math. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot of consent/upsell/cancel flows; source inspection for declarative button actions vs misleading controls. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise a form, submit invalid input, and observe timing and clarity of errors via a screenshot or a :user-invalid / aria-invalid probe. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source/DOM inspection for autocomplete attributes on form fields; a probe of autofill affordances. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: walkthrough checkout/subscription/auth/account flows when present; screenshot pricing, confirmation, cancellation and reauthentication states; inspect passkey/autocomplete support for sign-in and payment. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect transferred image bytes vs displayed size; source inspection for modern formats and resolution handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a long-task / network probe for background fetches and processing while idle or backgrounded. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a HAR summary shows third-party bytes, font/media weight and caching; screenshots/video reveal autoplay and decorative media; trace/layout evidence shows whether media/animation keeps work running. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for WebMCP / agentic-tool registration and agent-readable affordances. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for built-in AI (language model / summariser) usage. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT (not mandatory): compare heap snapshots for retained growth - a baseline, then one taken after repeating the interaction with `--interact` about 10x (the memory-tracer methodology: baseline -> repeat -> post -> compare). Performance.getMetrics (JSHeapUsedSize, Nodes) across the same before/after window is corroboration. If Chrome DevTools MCP is available, follow its memory-leak-debugging skill: capture baseline, target, and final snapshots, then use memlab or the provided comparison workflow rather than reading raw .heapsnapshot files directly. The package-native `heap` primitive remains the default path. This check is only meaningful where the page has a real interaction to repeat; for a static page with none, mark it not-applicable with a rationale rather than fabricating one. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a single `heap` summary's totals (nodeCount, totalSelfSizeBytes, constructor population) plus Performance.getMetrics (Nodes, JSHeapUsedSize) give the current footprint to judge against the page's purpose. Chrome DevTools MCP heap snapshots and memlab snapshot analysis can provide the same memory distribution when available. Read summaries or derived analysis, never raw snapshots unless a dedicated heap-analysis tool is doing the analysis. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the `heap` summary by constructor (Detached* nodes) compared across a before/after pair shows a growing detached-DOM population; an `evaluate` probe can sample listener/timer counts (e.g. getEventListeners-style counting, or instrumenting addEventListener/setInterval) before and after the repeated interaction to spot growth. Chrome DevTools MCP heap snapshots plus the memory-leak-debugging skill's common-leak guidance can corroborate detached DOM, listeners, closures, globals, and unbounded caches. Caveat from the memory-tracer and Chrome DevTools MCP guidance: detached nodes can be intentional caches, so judge confidence rather than asserting a bug. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT (not mandatory): a screenshot or computed background under an emulated prefers-color-scheme: dark condition will reveal whether surfaces re-tint; the page CSS / a color-scheme declaration is corroborating evidence. The model chooses the method.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video, or an in-page probe of getAnimations()/computed animation under an emulated prefers-reduced-motion: reduce condition, can show whether motion stops. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot under emulated prefers-contrast: more / forced-colors, or an axe/contrast probe, can show whether controls and text survive. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video of a route/state change shows whether it animates; the page source / ::view-transition usage corroborates. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source/CSS inspection for animation-timeline: scroll()/view(); a long-task / scroll-handler probe can flag the main-thread anti-pattern. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for scroll-snap / overscroll-behavior / physics-based easing vs custom pointermove listeners. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a transition video of scrolling, or CSS inspection for scroll-state container queries. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for anchor-name / position-anchor / position-try on overlays; a screenshot of an open overlay near a viewport edge can show drift. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for ::highlight / scroll-marker; a transition video can show whether attention is cued after navigation. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot on load, or a DOM probe for full-viewport overlays present before interaction. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: DOM/source inspection for popover / <dialog> / <details> vs custom overlay divs with manual dismiss handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot plus layout metrics can show the proportion of the viewport given to chrome vs content. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: layout metrics (scrollWidth vs innerWidth) and a screenshot at an emulated narrow mobile viewport reveal overflow. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: CSS inspection for @container / container-type; a computed-style probe of the same component in a wide vs narrow container shows whether it adapts. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a focus probe (focus an element, read the computed outline) or an axe target-size check; a screenshot of a focused control corroborates. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: screenshot the first viewport and key scrolled states; inspect heading structure, nav labels, button text, and visual hierarchy; a task walkthrough can show whether the next action is obvious. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: run the flow manually with screenshots/DOM snapshots at each step; compare expected vs actual path length; inspect form requirements, navigation continuity, and blockers. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise network delay/failure, invalid input, empty data and success states; screenshot the state messaging and recovery controls; inspect whether browser history and focus remain sensible. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse reports LCP/CLS/TBT directly and the model may run it; layout metrics + a layout-shift observer + a long-task observer (the evidence primitives) give the same signal first-party. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the layout primitive's CLS observer captures shifts; a transition video of the first seconds shows content jumping. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the layout primitive records long tasks; a heap summary shows the object population; Lighthouse reports TBT. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a HAR summary can reveal cache headers, redirects, render-blocking candidates, weight offenders and dependency shape; a trace/Lighthouse insight report can corroborate LCP discovery, render-blocking, font-display and document latency. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse/trace/code-coverage style evidence can flag unused JS/CSS, duplicated JavaScript and legacy code; a HAR summary shows third-party byte cost and request count. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe-core (injectable via the evaluate primitive) or Lighthouse's a11y audits enumerate these; a DOM probe of the accessibility-relevant attributes is a first-party alternative. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe contrast rules, a Lighthouse contrast audit, or an in-page probe computing contrast ratios from computed colours. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: axe/Lighthouse structural audits; a focus-walk probe (tab through, read activeElement + computed outline) is a first-party alternative. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot of body and heading text, plus CSS inspection for text-wrap / text alignment / font fallback handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse/axe target-size, meta-viewport and media-caption audits are useful signals; screenshots at narrow and zoomed conditions plus DOM/media inspection can corroborate. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: capture Runtime/Log CDP events, or a probe that reads collected errors; Lighthouse reports this too. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM/source probe for doctype/charset/img dimensions; CSS inspection for repetition; Lighthouse best-practices audits cover the rest. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: Lighthouse best-practices audits and DevTools inspector/deprecation signals can surface these; DOM/source probes can verify paste handlers and prompt timing. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe reads <title> and meta[name=description]; Lighthouse SEO audits cover the same ground. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe for anchor hrefs, viewport meta, and robots; Lighthouse SEO audits corroborate. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response status and headers, <link rel=canonical>, hreflang links, robots meta, robots.txt and sitemap.xml; Lighthouse SEO audits cover several of these. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect JSON-LD/microdata and social preview tags against visible content; Lighthouse has a manual structured-data audit, and ad-hoc probes can parse schema.org blocks. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response headers / page protocol via an evaluate probe or the network layer; Lighthouse best-practices flags HTTPS and CSP issues. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect network requests and third-party origins; a probe of analytics/beacon calls. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a probe for permission requests fired on load; source inspection for passkey / WebAuthn / navigator.credentials usage in auth flows. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect response headers and browser security state; Lighthouse/DevTools security audits can corroborate HSTS, clickjacking, Trusted Types, origin isolation and third-party cookie findings. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: load with scripting disabled or compare a no-JS fetch of the HTML against the rendered page; check for Baseline-aware fallbacks in source. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise menus near viewport edges with a screenshot; a probe of async/visibility behaviour. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a probe for a service worker registration and a web app manifest; test behaviour offline. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: simulate failed fetches/offline mode or inspect representative 404/500 routes; screenshots and DOM snapshots of error/loading/empty states show whether recovery is possible. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a DOM probe for <html lang>/dir and CSS inspection for logical vs physical properties. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for Intl.* usage vs hand-rolled formatting; a probe of rendered dates/numbers under a different locale. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for time-zone-aware date handling vs naive local Date math. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a screenshot of consent/upsell/cancel flows; source inspection for declarative button actions vs misleading controls. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: exercise a form, submit invalid input, and observe timing and clarity of errors via a screenshot or a :user-invalid / aria-invalid probe. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source/DOM inspection for autocomplete attributes on form fields; a probe of autofill affordances. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: walkthrough checkout/subscription/auth/account flows when present; screenshot pricing, confirmation, cancellation and reauthentication states; inspect passkey/autocomplete support for sign-in and payment. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: inspect transferred image bytes vs displayed size; source inspection for modern formats and resolution handling. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a long-task / network probe for background fetches and processing while idle or backgrounded. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a HAR summary shows third-party bytes, font/media weight and caching; screenshots/video reveal autoplay and decorative media; trace/layout evidence shows whether media/animation keeps work running. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for WebMCP / agentic-tool registration and agent-readable affordances. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: source inspection for built-in AI (language model / summariser) usage. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT (not mandatory): compare heap snapshots for retained growth - a baseline, then one taken after repeating the interaction with `--interact` about 10x (the memory-tracer methodology: baseline -> repeat -> post -> compare). Performance.getMetrics (JSHeapUsedSize, Nodes) across the same before/after window is corroboration. If Chrome DevTools MCP is available, follow its memory-leak-debugging skill: capture baseline, target, and final snapshots, then use memlab or the provided comparison workflow rather than reading raw .heapsnapshot files directly. The package-native `heap` primitive remains the default path. This check is only meaningful where the page has a real interaction to repeat; for a static page with none, mark it not-applicable with a rationale rather than fabricating one. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: a single `heap` summary's totals (nodeCount, totalSelfSizeBytes, constructor population) plus Performance.getMetrics (Nodes, JSHeapUsedSize) give the current footprint to judge against the page's purpose. Chrome DevTools MCP heap snapshots and memlab snapshot analysis can provide the same memory distribution when available. Read summaries or derived analysis, never raw snapshots unless a dedicated heap-analysis tool is doing the analysis. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT: the `heap` summary by constructor (Detached* nodes) compared across a before/after pair shows a growing detached-DOM population; an `evaluate` probe can sample listener/timer counts (e.g. getEventListeners-style counting, or instrumenting addEventListener/setInterval) before and after the repeated interaction to spot growth. Chrome DevTools MCP heap snapshots plus the memory-leak-debugging skill's common-leak guidance can corroborate detached DOM, listeners, closures, globals, and unbounded caches. Caveat from the memory-tracer and Chrome DevTools MCP guidance: detached nodes can be intentional caches, so judge confidence rather than asserting a bug. The model chooses.

Method used or attempted: No check-specific method was executed because the site report is unavailable.

Catalog test design: HINT (not mandatory): a screenshot or computed background under an emulated prefers-color-scheme: dark condition will reveal whether surfaces re-tint; the page CSS / a color-scheme declaration is corroborating evidence. The model chooses the method.

Method used or attempted: Compared permit-bound light and dark screenshots.

Catalog test design: HINT: a transition video, or an in-page probe of getAnimations()/computed animation under an emulated prefers-reduced-motion: reduce condition, can show whether motion stops. The model chooses.

Method used or attempted: Permit-bound evaluate probe under prefers-reduced-motion: reduce.

Catalog test design: HINT: a screenshot under emulated prefers-contrast: more / forced-colors, or an axe/contrast probe, can show whether controls and text survive. The model chooses.

Method used or attempted: Permit-bound forced-colors and prefers-contrast screenshot review.

Catalog test design: HINT: a transition video of a route/state change shows whether it animates; the page source / ::view-transition usage corroborates. The model chooses.

Method used or attempted: Reviewed the static entry and journey states.

Catalog test design: HINT: source/CSS inspection for animation-timeline: scroll()/view(); a long-task / scroll-handler probe can flag the main-thread anti-pattern. The model chooses.

Method used or attempted: Layout, reduced-motion probe, and journey state review.

Catalog test design: HINT: CSS inspection for scroll-snap / overscroll-behavior / physics-based easing vs custom pointermove listeners. The model chooses.

Method used or attempted: Reviewed DOM, layout, and bounded repeated native scrolling.

Catalog test design: HINT: a transition video of scrolling, or CSS inspection for scroll-state container queries. The model chooses.

Method used or attempted: Journey and layout review.

Catalog test design: HINT: CSS inspection for anchor-name / position-anchor / position-try on overlays; a screenshot of an open overlay near a viewport edge can show drift. The model chooses.

Method used or attempted: DOM review for popovers and anchored transient UI.

Catalog test design: HINT: CSS inspection for ::highlight / scroll-marker; a transition video can show whether attention is cued after navigation. The model chooses.

Method used or attempted: Visual hierarchy review of desktop, mobile, and journey baseline screenshots.

Catalog test design: HINT: a screenshot on load, or a DOM probe for full-viewport overlays present before interaction. The model chooses.

Method used or attempted: First-load screenshot review at the journey viewport.

Catalog test design: HINT: DOM/source inspection for popover / <dialog> / <details> vs custom overlay divs with manual dismiss handling. The model chooses.

Method used or attempted: DOM and role inspection.

Catalog test design: HINT: a screenshot plus layout metrics can show the proportion of the viewport given to chrome vs content. The model chooses.

Method used or attempted: Desktop and mobile screenshot review.

Catalog test design: HINT: layout metrics (scrollWidth vs innerWidth) and a screenshot at an emulated narrow mobile viewport reveal overflow. The model chooses.

Method used or attempted: Permit-bound 360x800 layout metrics and screenshot.

Catalog test design: HINT: CSS inspection for @container / container-type; a computed-style probe of the same component in a wide vs narrow container shows whether it adapts. The model chooses.

Method used or attempted: CSS feature probe and component inventory review.

Catalog test design: HINT: a focus probe (focus an element, read the computed outline) or an axe target-size check; a screenshot of a focused control corroborates. The model chooses.

Method used or attempted: Programmatically focused the first 20 focusable controls and inspected outlines and target rectangles.

Catalog test design: HINT: screenshot the first viewport and key scrolled states; inspect heading structure, nav labels, button text, and visual hierarchy; a task walkthrough can show whether the next action is obvious. The model chooses.

Method used or attempted: First viewport and semantic form review.

Catalog test design: HINT: run the flow manually with screenshots/DOM snapshots at each step; compare expected vs actual path length; inspect form requirements, navigation continuity, and blockers. The model chooses.

Method used or attempted: Reviewed the already-replayed strict journey and baseline screenshot.

Catalog test design: HINT: exercise network delay/failure, invalid input, empty data and success states; screenshot the state messaging and recovery controls; inspect whether browser history and focus remain sensible. The model chooses.

Method used or attempted: Consent state and journey ledger review.

Catalog test design: HINT: Lighthouse reports LCP/CLS/TBT directly and the model may run it; layout metrics + a layout-shift observer + a long-task observer (the evidence primitives) give the same signal first-party. The model chooses.

Method used or attempted: Permit-bound trace and layout observers.

Catalog test design: HINT: the layout primitive's CLS observer captures shifts; a transition video of the first seconds shows content jumping. The model chooses.

Method used or attempted: Layout shift observer during page load.

Catalog test design: HINT: the layout primitive records long tasks; a heap summary shows the object population; Lighthouse reports TBT. The model chooses.

Method used or attempted: Permit-bound performance trace.

Catalog test design: HINT: a HAR summary can reveal cache headers, redirects, render-blocking candidates, weight offenders and dependency shape; a trace/Lighthouse insight report can corroborate LCP discovery, render-blocking, font-display and document latency. The model chooses.

Method used or attempted: HAR summary plus rendered DOM review.

Catalog test design: HINT: Lighthouse/trace/code-coverage style evidence can flag unused JS/CSS, duplicated JavaScript and legacy code; a HAR summary shows third-party byte cost and request count. The model chooses.

Method used or attempted: HAR resource-type and page-complexity comparison.

Catalog test design: HINT: axe-core (injectable via the evaluate primitive) or Lighthouse's a11y audits enumerate these; a DOM probe of the accessibility-relevant attributes is a first-party alternative. The model chooses.

Method used or attempted: DOM accessibility semantics and visible control review.

Catalog test design: HINT: axe contrast rules, a Lighthouse contrast audit, or an in-page probe computing contrast ratios from computed colours. The model chooses.

Method used or attempted: Normal dark and forced-colors screenshot review.

Catalog test design: HINT: axe/Lighthouse structural audits; a focus-walk probe (tab through, read activeElement + computed outline) is a first-party alternative. The model chooses.

Method used or attempted: Heading, landmark, and programmatic focus inspection.

Catalog test design: HINT: a screenshot of body and heading text, plus CSS inspection for text-wrap / text alignment / font fallback handling. The model chooses.

Method used or attempted: Desktop, mobile, and contrast screenshot review.

Catalog test design: HINT: Lighthouse/axe target-size, meta-viewport and media-caption audits are useful signals; screenshots at narrow and zoomed conditions plus DOM/media inspection can corroborate. The model chooses.

Method used or attempted: Narrow viewport layout and screenshot review.

Catalog test design: HINT: capture Runtime/Log CDP events, or a probe that reads collected errors; Lighthouse reports this too. The model chooses.

Method used or attempted: Observed permit-bound Chrome CLI load, trace completion, and journey event ledger.

Catalog test design: HINT: a DOM/source probe for doctype/charset/img dimensions; CSS inspection for repetition; Lighthouse best-practices audits cover the rest. The model chooses.

Method used or attempted: DOM, document metadata, image, and layout inspection.

Catalog test design: HINT: Lighthouse best-practices audits and DevTools inspector/deprecation signals can surface these; DOM/source probes can verify paste handlers and prompt timing. The model chooses.

Method used or attempted: DOM and runtime surface review.

Catalog test design: HINT: a DOM probe reads <title> and meta[name=description]; Lighthouse SEO audits cover the same ground. The model chooses.

Method used or attempted: DOM metadata probe.

Catalog test design: HINT: a DOM probe for anchor hrefs, viewport meta, and robots; Lighthouse SEO audits corroborate. The model chooses.

Method used or attempted: DOM links, raw HTML comparison, and mobile layout inspection.

Catalog test design: HINT: inspect response status and headers, <link rel=canonical>, hreflang links, robots meta, robots.txt and sitemap.xml; Lighthouse SEO audits cover several of these. The model chooses.

Method used or attempted: Discoverability fetch and metadata inspection.

Catalog test design: HINT: inspect JSON-LD/microdata and social preview tags against visible content; Lighthouse has a manual structured-data audit, and ad-hoc probes can parse schema.org blocks. The model chooses.

Method used or attempted: Entity and metadata review.

Catalog test design: HINT: inspect response headers / page protocol via an evaluate probe or the network layer; Lighthouse best-practices flags HTTPS and CSP issues. The model chooses.

Method used or attempted: Reviewed transport state, baseline security-header presence, and cookie-attribute counts from retained private evidence.

Catalog test design: HINT: inspect network requests and third-party origins; a probe of analytics/beacon calls. The model chooses.

Method used or attempted: Reviewed categorical request-origin and cookie-attribute counts from retained private evidence.

Catalog test design: HINT: a probe for permission requests fired on load; source inspection for passkey / WebAuthn / navigator.credentials usage in auth flows. The model chooses.

Method used or attempted: Initial-load prompt and auth-surface review.

Catalog test design: HINT: inspect response headers and browser security state; Lighthouse/DevTools security audits can corroborate HSTS, clickjacking, Trusted Types, origin isolation and third-party cookie findings. The model chooses.

Method used or attempted: Reviewed baseline browser-policy header presence from retained private evidence.

Catalog test design: HINT: load with scripting disabled or compare a no-JS fetch of the HTML against the rendered page; check for Baseline-aware fallbacks in source. The model chooses.

Method used or attempted: Permit-bound discoverability raw fetch and crawler screenshot.

Catalog test design: HINT: exercise menus near viewport edges with a screenshot; a probe of async/visibility behaviour. The model chooses.

Method used or attempted: Journey viewport screenshot and dialog rectangle review.

Catalog test design: HINT: a probe for a service worker registration and a web app manifest; test behaviour offline. The model chooses.

Method used or attempted: Manifest and service-worker registration probe.

Catalog test design: HINT: simulate failed fetches/offline mode or inspect representative 404/500 routes; screenshots and DOM snapshots of error/loading/empty states show whether recovery is possible. The model chooses.

Method used or attempted: Site purpose and bounded policy review.

Catalog test design: HINT: a DOM probe for <html lang>/dir and CSS inspection for logical vs physical properties. The model chooses.

Method used or attempted: DOM and stylesheet feature probe.

Catalog test design: HINT: source inspection for Intl.* usage vs hand-rolled formatting; a probe of rendered dates/numbers under a different locale. The model chooses.

Method used or attempted: Rendered content and control inventory review.

Catalog test design: HINT: source inspection for time-zone-aware date handling vs naive local Date math. The model chooses.

Method used or attempted: Rendered content and platform probe.

Catalog test design: HINT: a screenshot of consent/upsell/cancel flows; source inspection for declarative button actions vs misleading controls. The model chooses.

Method used or attempted: Consent wall visual and copy review.

Catalog test design: HINT: exercise a form, submit invalid input, and observe timing and clarity of errors via a screenshot or a :user-invalid / aria-invalid probe. The model chooses.

Method used or attempted: Form requirements inspection.

Catalog test design: HINT: source/DOM inspection for autocomplete attributes on form fields; a probe of autofill affordances. The model chooses.

Method used or attempted: Search control semantics inspection.

Catalog test design: HINT: walkthrough checkout/subscription/auth/account flows when present; screenshot pricing, confirmation, cancellation and reauthentication states; inspect passkey/autocomplete support for sign-in and payment. The model chooses.

Method used or attempted: Consent choice visual review.

Catalog test design: HINT: inspect transferred image bytes vs displayed size; source inspection for modern formats and resolution handling. The model chooses.

Method used or attempted: Image and HAR resource review.

Catalog test design: HINT: a long-task / network probe for background fetches and processing while idle or backgrounded. The model chooses.

Method used or attempted: HAR and page-complexity review.

Catalog test design: HINT: a HAR summary shows third-party bytes, font/media weight and caching; screenshots/video reveal autoplay and decorative media; trace/layout evidence shows whether media/animation keeps work running. The model chooses.

Method used or attempted: HAR and tracker review.

Catalog test design: HINT: source inspection for WebMCP / agentic-tool registration and agent-readable affordances. The model chooses.

Method used or attempted: Raw HTML and semantic form inspection.

Catalog test design: HINT: source inspection for built-in AI (language model / summariser) usage. The model chooses.

Method used or attempted: Page purpose and runtime capability review.

Catalog test design: HINT (not mandatory): compare heap snapshots for retained growth - a baseline, then one taken after repeating the interaction with `--interact` about 10x (the memory-tracer methodology: baseline -> repeat -> post -> compare). Performance.getMetrics (JSHeapUsedSize, Nodes) across the same before/after window is corroboration. If Chrome DevTools MCP is available, follow its memory-leak-debugging skill: capture baseline, target, and final snapshots, then use memlab or the provided comparison workflow rather than reading raw .heapsnapshot files directly. The package-native `heap` primitive remains the default path. This check is only meaningful where the page has a real interaction to repeat; for a static page with none, mark it not-applicable with a rationale rather than fabricating one. The model chooses.

Method used or attempted: Compared independent heap summaries before and after ten bounded native scroll down/up cycles.

Catalog test design: HINT: a single `heap` summary's totals (nodeCount, totalSelfSizeBytes, constructor population) plus Performance.getMetrics (Nodes, JSHeapUsedSize) give the current footprint to judge against the page's purpose. Chrome DevTools MCP heap snapshots and memlab snapshot analysis can provide the same memory distribution when available. Read summaries or derived analysis, never raw snapshots unless a dedicated heap-analysis tool is doing the analysis. The model chooses.

Method used or attempted: Heap summary and DOM-size review.

Catalog test design: HINT: the `heap` summary by constructor (Detached* nodes) compared across a before/after pair shows a growing detached-DOM population; an `evaluate` probe can sample listener/timer counts (e.g. getEventListeners-style counting, or instrumenting addEventListener/setInterval) before and after the repeated interaction to spot growth. Chrome DevTools MCP heap snapshots plus the memory-leak-debugging skill's common-leak guidance can corroborate detached DOM, listeners, closures, globals, and unbounded caches. Caveat from the memory-tracer and Chrome DevTools MCP guidance: detached nodes can be intentional caches, so judge confidence rather than asserting a bug. The model chooses.

Method used or attempted: Compared heap constructor summaries across repeated bounded scroll.

Catalog test design: HINT (not mandatory): a screenshot or computed background under an emulated prefers-color-scheme: dark condition will reveal whether surfaces re-tint; the page CSS / a color-scheme declaration is corroborating evidence. The model chooses the method.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a transition video, or an in-page probe of getAnimations()/computed animation under an emulated prefers-reduced-motion: reduce condition, can show whether motion stops. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a screenshot under emulated prefers-contrast: more / forced-colors, or an axe/contrast probe, can show whether controls and text survive. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a transition video of a route/state change shows whether it animates; the page source / ::view-transition usage corroborates. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: source/CSS inspection for animation-timeline: scroll()/view(); a long-task / scroll-handler probe can flag the main-thread anti-pattern. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: CSS inspection for scroll-snap / overscroll-behavior / physics-based easing vs custom pointermove listeners. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a transition video of scrolling, or CSS inspection for scroll-state container queries. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: CSS inspection for anchor-name / position-anchor / position-try on overlays; a screenshot of an open overlay near a viewport edge can show drift. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: CSS inspection for ::highlight / scroll-marker; a transition video can show whether attention is cued after navigation. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a screenshot on load, or a DOM probe for full-viewport overlays present before interaction. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: DOM/source inspection for popover / <dialog> / <details> vs custom overlay divs with manual dismiss handling. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a screenshot plus layout metrics can show the proportion of the viewport given to chrome vs content. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: layout metrics (scrollWidth vs innerWidth) and a screenshot at an emulated narrow mobile viewport reveal overflow. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: CSS inspection for @container / container-type; a computed-style probe of the same component in a wide vs narrow container shows whether it adapts. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a focus probe (focus an element, read the computed outline) or an axe target-size check; a screenshot of a focused control corroborates. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: screenshot the first viewport and key scrolled states; inspect heading structure, nav labels, button text, and visual hierarchy; a task walkthrough can show whether the next action is obvious. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: run the flow manually with screenshots/DOM snapshots at each step; compare expected vs actual path length; inspect form requirements, navigation continuity, and blockers. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: exercise network delay/failure, invalid input, empty data and success states; screenshot the state messaging and recovery controls; inspect whether browser history and focus remain sensible. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: Lighthouse reports LCP/CLS/TBT directly and the model may run it; layout metrics + a layout-shift observer + a long-task observer (the evidence primitives) give the same signal first-party. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: the layout primitive's CLS observer captures shifts; a transition video of the first seconds shows content jumping. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: the layout primitive records long tasks; a heap summary shows the object population; Lighthouse reports TBT. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a HAR summary can reveal cache headers, redirects, render-blocking candidates, weight offenders and dependency shape; a trace/Lighthouse insight report can corroborate LCP discovery, render-blocking, font-display and document latency. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: Lighthouse/trace/code-coverage style evidence can flag unused JS/CSS, duplicated JavaScript and legacy code; a HAR summary shows third-party byte cost and request count. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: axe-core (injectable via the evaluate primitive) or Lighthouse's a11y audits enumerate these; a DOM probe of the accessibility-relevant attributes is a first-party alternative. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: axe contrast rules, a Lighthouse contrast audit, or an in-page probe computing contrast ratios from computed colours. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: axe/Lighthouse structural audits; a focus-walk probe (tab through, read activeElement + computed outline) is a first-party alternative. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a screenshot of body and heading text, plus CSS inspection for text-wrap / text alignment / font fallback handling. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: Lighthouse/axe target-size, meta-viewport and media-caption audits are useful signals; screenshots at narrow and zoomed conditions plus DOM/media inspection can corroborate. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: capture Runtime/Log CDP events, or a probe that reads collected errors; Lighthouse reports this too. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: a DOM/source probe for doctype/charset/img dimensions; CSS inspection for repetition; Lighthouse best-practices audits cover the rest. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: Lighthouse best-practices audits and DevTools inspector/deprecation signals can surface these; DOM/source probes can verify paste handlers and prompt timing. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a DOM probe reads <title> and meta[name=description]; Lighthouse SEO audits cover the same ground. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a DOM probe for anchor hrefs, viewport meta, and robots; Lighthouse SEO audits corroborate. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: inspect response status and headers, <link rel=canonical>, hreflang links, robots meta, robots.txt and sitemap.xml; Lighthouse SEO audits cover several of these. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: inspect JSON-LD/microdata and social preview tags against visible content; Lighthouse has a manual structured-data audit, and ad-hoc probes can parse schema.org blocks. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: inspect response headers / page protocol via an evaluate probe or the network layer; Lighthouse best-practices flags HTTPS and CSP issues. The model chooses.

Method used or attempted: Reviewed transport state, baseline security-header presence, and cookie-attribute counts from retained private evidence.

Catalog test design: HINT: inspect network requests and third-party origins; a probe of analytics/beacon calls. The model chooses.

Method used or attempted: Reviewed categorical request-origin and cookie-attribute counts from retained private evidence.

Catalog test design: HINT: a probe for permission requests fired on load; source inspection for passkey / WebAuthn / navigator.credentials usage in auth flows. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: inspect response headers and browser security state; Lighthouse/DevTools security audits can corroborate HSTS, clickjacking, Trusted Types, origin isolation and third-party cookie findings. The model chooses.

Method used or attempted: Reviewed baseline browser-policy header presence from retained private evidence.

Catalog test design: HINT: load with scripting disabled or compare a no-JS fetch of the HTML against the rendered page; check for Baseline-aware fallbacks in source. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: exercise menus near viewport edges with a screenshot; a probe of async/visibility behaviour. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a probe for a service worker registration and a web app manifest; test behaviour offline. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: simulate failed fetches/offline mode or inspect representative 404/500 routes; screenshots and DOM snapshots of error/loading/empty states show whether recovery is possible. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: a DOM probe for <html lang>/dir and CSS inspection for logical vs physical properties. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: source inspection for Intl.* usage vs hand-rolled formatting; a probe of rendered dates/numbers under a different locale. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: source inspection for time-zone-aware date handling vs naive local Date math. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a screenshot of consent/upsell/cancel flows; source inspection for declarative button actions vs misleading controls. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: exercise a form, submit invalid input, and observe timing and clarity of errors via a screenshot or a :user-invalid / aria-invalid probe. The model chooses.

Method used or attempted: Permit-bound recon plus strict journey replay result

Catalog test design: HINT: source/DOM inspection for autocomplete attributes on form fields; a probe of autofill affordances. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: walkthrough checkout/subscription/auth/account flows when present; screenshot pricing, confirmation, cancellation and reauthentication states; inspect passkey/autocomplete support for sign-in and payment. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: inspect transferred image bytes vs displayed size; source inspection for modern formats and resolution handling. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a long-task / network probe for background fetches and processing while idle or backgrounded. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a HAR summary shows third-party bytes, font/media weight and caching; screenshots/video reveal autoplay and decorative media; trace/layout evidence shows whether media/animation keeps work running. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: source inspection for WebMCP / agentic-tool registration and agent-readable affordances. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: source inspection for built-in AI (language model / summariser) usage. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT (not mandatory): compare heap snapshots for retained growth - a baseline, then one taken after repeating the interaction with `--interact` about 10x (the memory-tracer methodology: baseline -> repeat -> post -> compare). Performance.getMetrics (JSHeapUsedSize, Nodes) across the same before/after window is corroboration. If Chrome DevTools MCP is available, follow its memory-leak-debugging skill: capture baseline, target, and final snapshots, then use memlab or the provided comparison workflow rather than reading raw .heapsnapshot files directly. The package-native `heap` primitive remains the default path. This check is only meaningful where the page has a real interaction to repeat; for a static page with none, mark it not-applicable with a rationale rather than fabricating one. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: a single `heap` summary's totals (nodeCount, totalSelfSizeBytes, constructor population) plus Performance.getMetrics (Nodes, JSHeapUsedSize) give the current footprint to judge against the page's purpose. Chrome DevTools MCP heap snapshots and memlab snapshot analysis can provide the same memory distribution when available. Read summaries or derived analysis, never raw snapshots unless a dedicated heap-analysis tool is doing the analysis. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT: the `heap` summary by constructor (Detached* nodes) compared across a before/after pair shows a growing detached-DOM population; an `evaluate` probe can sample listener/timer counts (e.g. getEventListeners-style counting, or instrumenting addEventListener/setInterval) before and after the repeated interaction to spot growth. Chrome DevTools MCP heap snapshots plus the memory-leak-debugging skill's common-leak guidance can corroborate detached DOM, listeners, closures, globals, and unbounded caches. Caveat from the memory-tracer and Chrome DevTools MCP guidance: detached nodes can be intentional caches, so judge confidence rather than asserting a bug. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, trace, HAR, security, discoverability, or heap evidence as applicable

Catalog test design: HINT (not mandatory): a screenshot or computed background under an emulated prefers-color-scheme: dark condition will reveal whether surfaces re-tint; the page CSS / a color-scheme declaration is corroborating evidence. The model chooses the method.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: a transition video, or an in-page probe of getAnimations()/computed animation under an emulated prefers-reduced-motion: reduce condition, can show whether motion stops. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: a screenshot under emulated prefers-contrast: more / forced-colors, or an axe/contrast probe, can show whether controls and text survive. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: a transition video of a route/state change shows whether it animates; the page source / ::view-transition usage corroborates. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: source/CSS inspection for animation-timeline: scroll()/view(); a long-task / scroll-handler probe can flag the main-thread anti-pattern. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: CSS inspection for scroll-snap / overscroll-behavior / physics-based easing vs custom pointermove listeners. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: a transition video of scrolling, or CSS inspection for scroll-state container queries. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: CSS inspection for anchor-name / position-anchor / position-try on overlays; a screenshot of an open overlay near a viewport edge can show drift. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: CSS inspection for ::highlight / scroll-marker; a transition video can show whether attention is cued after navigation. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: a screenshot on load, or a DOM probe for full-viewport overlays present before interaction. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: DOM/source inspection for popover / <dialog> / <details> vs custom overlay divs with manual dismiss handling. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: a screenshot plus layout metrics can show the proportion of the viewport given to chrome vs content. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: layout metrics (scrollWidth vs innerWidth) and a screenshot at an emulated narrow mobile viewport reveal overflow. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: CSS inspection for @container / container-type; a computed-style probe of the same component in a wide vs narrow container shows whether it adapts. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: a focus probe (focus an element, read the computed outline) or an axe target-size check; a screenshot of a focused control corroborates. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: screenshot the first viewport and key scrolled states; inspect heading structure, nav labels, button text, and visual hierarchy; a task walkthrough can show whether the next action is obvious. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: run the flow manually with screenshots/DOM snapshots at each step; compare expected vs actual path length; inspect form requirements, navigation continuity, and blockers. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: exercise network delay/failure, invalid input, empty data and success states; screenshot the state messaging and recovery controls; inspect whether browser history and focus remain sensible. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: Lighthouse reports LCP/CLS/TBT directly and the model may run it; layout metrics + a layout-shift observer + a long-task observer (the evidence primitives) give the same signal first-party. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: the layout primitive's CLS observer captures shifts; a transition video of the first seconds shows content jumping. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: the layout primitive records long tasks; a heap summary shows the object population; Lighthouse reports TBT. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: a HAR summary can reveal cache headers, redirects, render-blocking candidates, weight offenders and dependency shape; a trace/Lighthouse insight report can corroborate LCP discovery, render-blocking, font-display and document latency. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: Lighthouse/trace/code-coverage style evidence can flag unused JS/CSS, duplicated JavaScript and legacy code; a HAR summary shows third-party byte cost and request count. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: axe-core (injectable via the evaluate primitive) or Lighthouse's a11y audits enumerate these; a DOM probe of the accessibility-relevant attributes is a first-party alternative. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: axe contrast rules, a Lighthouse contrast audit, or an in-page probe computing contrast ratios from computed colours. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: axe/Lighthouse structural audits; a focus-walk probe (tab through, read activeElement + computed outline) is a first-party alternative. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: a screenshot of body and heading text, plus CSS inspection for text-wrap / text alignment / font fallback handling. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: Lighthouse/axe target-size, meta-viewport and media-caption audits are useful signals; screenshots at narrow and zoomed conditions plus DOM/media inspection can corroborate. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: capture Runtime/Log CDP events, or a probe that reads collected errors; Lighthouse reports this too. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: a DOM/source probe for doctype/charset/img dimensions; CSS inspection for repetition; Lighthouse best-practices audits cover the rest. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: Lighthouse best-practices audits and DevTools inspector/deprecation signals can surface these; DOM/source probes can verify paste handlers and prompt timing. The model chooses.

Method used or attempted: Attempted bounded entry-page inspection and/or strict pre-replayed journey.

Catalog test design: HINT: a DOM probe reads <title> and meta[name=description]; Lighthouse SEO audits cover the same ground. The model chooses.

Method used or attempted: Permit-bound evidence review for the entry state.

Catalog test design: HINT: a DOM probe for anchor hrefs, viewport meta, and robots; Lighthouse SEO audits corroborate. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: inspect response status and headers, <link rel=canonical>, hreflang links, robots meta, robots.txt and sitemap.xml; Lighthouse SEO audits cover several of these. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.

Catalog test design: HINT: inspect JSON-LD/microdata and social preview tags against visible content; Lighthouse has a manual structured-data audit, and ad-hoc probes can parse schema.org blocks. The model chooses.

Method used or attempted: Permit-bound screenshot, DOM/evaluate, layout, HAR, discoverability, security, image, or journey evidence as cited.