Astro Mobile Safari Testing and Release Checklist

This checklist is mandatory before production. Any unchecked required item blocks release. PageSpeed must report 100 for Performance, Accessibility, Best Practices, and SEO on both mobile and desktop.

Project onboarding and service decisions

• PROJECT-ONBOARDING.md was read before implementation began.
• templates/project-onboarding.md was copied into the target repository and completed.
• Every selected workflow identifies the inputs and evidence required to perform it.
• Every external source is classified as required, conditional, optional, not used, or blocked.
• The project owner explicitly decided which third-party paid services may be purchased or connected.
• Every approved external account has a named owner, least-privilege scope, masked access check, and fallback or blocker.
• Ahrefs is recorded as optional unless a reviewed project contract requires it for one specific workflow.
• The active build and test operating system is recorded.
• A qualified macOS runner with full Xcode, an iOS runtime, a Simulator device name, and a UDID is assigned for native Safari testing.
• If the active operator uses Windows or Linux, the exact candidate handoff to the qualified Mac runner is documented.
• Production is blocked when required native iOS Simulator evidence cannot be produced.

Stanford Rule content quality

• STANFORD-RULE-CONTENT-QUALITY.md was read before public content was written or revised.
• Every public route names its intended audience and the task that audience is trying to complete.
• The opening addresses the reader’s need before internal implementation details.
• Requirements, optional tools, examples, and limitations are clearly distinguished.
• Unfamiliar technical terms are defined before they carry an argument.
• Machine-like filler, inflated marketing language, and repetitive sentence frames were removed.
• Sentence length, paragraph length, and reading ease pass the reviewed route thresholds.
• A reviewer used the senior psychology professor perspective to assess approachability, human tone, clear purpose, and evidence awareness.
• Every editorial review has a specific note, current date, audience match, and exact built-content hash.
• The content quality verifier passes against the exact production build.
• No result is described as AI detection or as a Stanford University policy.

Baseline

• Fetch the configured Go for Launch upstream and confirm the checkout is current.
• Record the toolkit repository, branch, and commit in release evidence.
• Read applicable repository instructions.
• Record the existing working tree and preserve unrelated changes.
• Confirm the real Astro application root.
• Inventory every public and intentionally private route.
• Capture legacy desktop, tablet, and mobile references.
• Record the canonical hostname, staging target, and deployment contract.

Project-Controlled Design-System Conformance

• Read Configurable Design-System Gate and record the project mode, framework, scope, and reviewer policy.
• Run the design gate and preserve its machine-readable result, including when the mode is off.
• Confirm the design setting does not disable any mandatory core gate.
• For advisory or required review when the configured scope applies, read Design Optimization and Brand Continuity and complete templates/design-optimization-brief.md.
• Treat findings as production blockers only when the mode is required.
• Do not claim conformance without an applicable passing review.

Complete the remaining items in this section only when design review is applicable:

• Record the approved brand source and design authority.
• Inventory logo, color, typography, imagery, shape language, voice, composition, and trust anchors.
• Record whether each brand anchor will be preserved, refined, replaced with approval, or remains unknown.
• Diagnose hierarchy, density, interaction, and responsive anatomy before changing styles.
• Classify every bar near the logo or navigation as an alert, announcement, utility bar, or brand statement.
• Evergreen positioning and marketing copy are not styled as alerts.
• Every announcement has a content owner, destination when needed, and review or expiration date.
• The header uses one primary control layer unless a second layer has a distinct documented user purpose.
• Status dots, warning colors, and alert icons represent genuine current states.
• Capture baseline and candidate first-viewport and full-page evidence at 1440 by 1000, 1024 by 900, 768 by 1024, 390 by 844, and 320 CSS pixels wide.
• Review the homepage and every affected route family, including long-form pages and open interaction states.
• Page purpose and the primary action are clear at every required viewport.
• Hero scale supports the page purpose and does not delay useful content.
• Heading measure, content density, section rhythm, grid transitions, and card anatomy pass visual review.
• The site remains recognizable without relying only on the logo.
• Glass, translucency, blur, elevation, and motion support hierarchy and remain legible.
• Text, logos, and interface icons remain sharp at native capture resolution.
• Blur is isolated to decorative layers and is not applied directly to content-bearing containers.
• Changed components include all applicable default, hover, focus, active, disabled, loading, success, and error states.
• Visual acceptance is recorded for the exact candidate.
• A designated human stakeholder approved the exact candidate when reviewerRequired is true.

Static Content and First Paint

• Primary content is present with JavaScript disabled.
• No startup class hides html, body, main, or a full-page wrapper.
• Critical styles are available before first paint.
• Fonts have stable fallbacks.
• Images have intrinsic dimensions.
• The first mobile viewport is useful and not blank.
• Desktop does not request mobile-only first-viewport artwork.
• Mobile does not request desktop-only first-viewport artwork.
• Hidden responsive resource variants are not downloaded.
• The declared preload and fetch priority match the browser’s measured LCP resource.

Layout

• No horizontal overflow at 390, 375, 360, and 320 CSS pixels.
• Fixed navigation clears the first heading.
• Same-page anchor targets clear fixed navigation.
• Images and logos preserve source aspect ratios.
• Cards, grids, and fixed-format controls have stable dimensions.
• Long text wraps without expanding the document width.
• Modals fit the visual viewport and scroll internally when needed.

Safari Rendering

• No blanket transforms or hidden backfaces on text.
• No permanent broad will-change declarations.
• No viewport-sized live blur filters.
• No unused anti-flicker rule can hide content.
• Long pages scroll from top to bottom and back without stalls.
• No uncaught JavaScript page errors occur.

Navigation and Interaction

• Mobile menu opens and closes.
• Direct top-level navigation links work with touch.
• Every dropdown opens with touch.
• At least one destination inside every dropdown navigates with touch.
• Dropdown cleanup does not hide the tapped link before default navigation.
• Escape and keyboard focus behavior work where applicable.
• Modals lock background scrolling only while open.
• Closing every modal restores page scrolling.
• Smooth scrolling respects reduced motion.
• Forms validate and submit on desktop and mobile.
• Every persistent side rail, table of contents, policy rail, and vertical tab list is marked with data-side-navigation.
• Every side-navigation item is a real link with a valid native destination when JavaScript is unavailable.
• Every side-navigation item is activated in browser tests, with its destination, active state, target visibility, keyboard behavior, and touch behavior verified.

Automated Tests

• Astro diagnostics pass.
• Production build passes.
• The mandatory render sharpness gate passes against the exact built candidate and its JSON report is preserved.
• The side-navigation verifier passes against the exact built candidate and its JSON report is preserved.
• Every CSS illustration, diagram, generated page graphic, chart, hero artwork, and website email graphic in scope is marked for visual composition verification.
• The visual composition verifier passes in Chromium and WebKit at desktop, mobile, and 320 CSS pixel viewports.
• Marked labels remain inside their artboards, do not overlap each other, and are not crossed by decorative geometry.
• Reviewed horizontal and vertical fill thresholds reject accidental dead zones without forcing decorative density.
• Every generated artboard capture is reviewed at native size for reading order, balance, useful empty space, connector clarity, and professional craft.
• The visual composition JSON report, screenshots, reviewer, and decision are preserved with the exact candidate evidence.
• Every named first-choice font is shipped or replaced with an approved system stack.
• Inline logos and interface SVGs avoid accidental fractional view-box scaling.
• The normal production build generates /sitemap.xml and runs the sitemap verifier.
• Every indexable built canonical appears exactly once in the sitemap.
• Every sitemap URL maps to an indexable built page on the canonical origin.
• robots.txt advertises the exact canonical sitemap URL.
• Unit and server-side tests pass.
• Chromium desktop tests pass.
• Chromium mobile tests pass.
• Playwright WebKit tests pass using tap() for touch flows.
• Browser request traces cover every breakpoint where the first-viewport resource set changes.
• The LCP element, asset URL, request start, preload, and responsive source are asserted in browser tests.
• Route matrix has no missing pages or required assets.
• Static-output SEO validator passes for titles, descriptions, canonicals, Open Graph, JSON-LD, and headings.
• Site-health audit passes against final HTML, CSS, images, metadata, internal links, redirects, and robots.txt.
• Semantic SEO audit passes against canonicals, titles, page intent, content depth, and citations.
• Every indexable route is covered by a reviewed page-intent rule when complete coverage is required.
• Titles are descriptive, complete, within the reviewed editorial budget, and aligned with the visible h1 and primary content.
• Citation-bearing content uses descriptive anchors and valid HTTP or HTTPS source URLs.
• Required citation evidence records include claim terms, source terms, reviewer, review date, and support limitations.
• The machine-readable semantic SEO report is preserved with release evidence.
• Every referenced local image is present and within the reviewed byte budget.
• Internal links point directly to built canonical routes without a redirect hop.
• Every indexable page has an incoming internal link, or a documented approved exception.
• Titles and descriptions are unique and within the configured length limits.
• The machine-readable site-health report is preserved with release evidence.
• Open Graph contact sheets include every indexable page and were reviewed at full size.
• Open Graph cards have no overlapping or clipped text, jagged or upscaled artwork, unintended transparency, incorrect page content, or unsafe crop placement.
• Every social-card source image is visually informative, with flat placeholders and empty exports replaced by an approved designed fallback.
• The normal build reused existing approved Open Graph cards without changing their bytes, names, encoding, or modification times.
• Any regenerated card was explicitly requested because a rendering input changed, and unchanged cards were not rewritten.
• Card input fingerprints exclude unrelated SEO policy, sitemap, citation, dependency, timestamp, environment, and build values.
• Displayed destinations contain no ellipsis or unusable truncation, supporting text remains readable, and letters with descenders are not clipped.
• Every card declares its intended sharing purpose and passes a named human readability assessment for that purpose.
• Every color and type family is brand approved, padding meets the brand safe-area minimum, and visual hierarchy preserves brand integrity.
• The authoritative brand guide and current brand kit were reviewed before selecting each logo, mark, wordmark, icon, seal, or branded illustration.
• Every brand asset matches its recorded SHA-256, named variant, allowed surface, intrinsic aspect ratio, minimum size, and clear-space requirement.
• Light, dark, colored, patterned, and photographic surfaces use the exact variants approved for those contexts. No mark was cropped from another lockup or recolored for convenience.
• Required contact information is visible, useful, accurate, and readable. Cards that do not require contact information explicitly record that decision.
• Font sizes remain within reviewed minimum and maximum limits, text regions do not overlap, and no glyph ink reaches a clipping boundary.
• The hash-bound Open Graph approval manifest matches both the rendering input hashes and exact image hashes in the candidate.
• Every indexable page has exactly one h1, and heading levels do not jump.
• Image output validator passes for alternative text, intrinsic dimensions, responsive srcset, sizes, and picture fallbacks.
• Localized sites have complete reciprocal hreflang clusters, localized self-canonicals, and x-default.
• Visual regression results are recorded, and they pass when the design mode is required.
• Form tests use mocked delivery except for one controlled staging submission.

Native Simulator

• Full Xcode and an iOS runtime are installed.
• One target iPhone UDID is recorded.
• Unexpected booted devices are shut down.
• Simulator is opened with the target UDID.
• Local production output is tested in Safari.
• Staging is tested in Safari.
• Mobile menu dropdown destinations are tapped and confirmed.
• Long-page scrolling is inspected.
• Modal open and close behavior is inspected.
• Address bar and visible destination content confirm navigation.

Performance, SEO, and Accessibility

• CLOUDFLARE-OBSERVABILITY.md was read for a Cloudflare-hosted project.
• Approved Cloudflare account analytics access was checked with a masked least-privilege request, or its unavailable state was recorded.
• An advisory production RUM baseline records the exact hostname, time window, sample minimum, thresholds, route and device groups, and LCP debug elements when access exists.
• Cloudflare RUM evidence is not substituted for PageSpeed, WebKit, or native Safari evidence.
• PageSpeed mobile Performance equals 100.
• PageSpeed mobile Accessibility equals 100.
• PageSpeed mobile Best Practices equals 100.
• PageSpeed mobile SEO equals 100.
• PageSpeed desktop Performance equals 100.
• PageSpeed desktop Accessibility equals 100.
• PageSpeed desktop Best Practices equals 100.
• PageSpeed desktop SEO equals 100.
• Normal text and large text meet WCAG 2.2 AA contrast requirements in every changed surface state.
• Changed content and controls remain usable at 200 percent text resize.
• Ordinary page content reflows without two-dimensional scrolling at 320 CSS pixels.
• WCAG text spacing overrides do not clip, overlap, or hide content.
• Pointer target size and spacing pass, with important touch controls targeting 44 by 44 CSS pixels where practical.
• Reduced motion, increased contrast, and reduced transparency behavior pass where the platform exposes those preferences.
• Canonical metadata is correct per page.
• Open Graph previews use the correct per-page content and image.
• Every indexable public page declares its own unique Open Graph image. No shared fallback image across pages.
• Each declared Open Graph image returns HTTP 200, a raster content type, and the exact declared dimensions.
• Representative cards render correctly in a real messaging or social preview, including the application-provided title and description outside the image.
• Structured data parses and matches visible content.
• Every JSON-LD root has @context and @type, or a typed @graph.
• Answer-focused pages use complete natural-language question headings and immediate, self-contained answers.
• FAQ answers are present in built HTML and remain readable without JavaScript.
• FAQ or Q&A structured data is used only for eligible visible content and exactly matches that content.
• Technical, regulatory, legal, statistical, and time-sensitive claims cite current primary sources.
• A last-reviewed date and responsible reviewer are visible on maintained guidance.
• SEO content work records credible question evidence from sources available to the project, with Search Console or optional Ahrefs evidence only when approved access exists.
• Public discovery files include only intended public routes.
• The staged /sitemap.xml and every referenced child sitemap return HTTP 200 with XML content.
• Search Console property access and permission were checked when approved access exists.
• The canonical sitemap is submitted in Search Console, or the exact access or verification blocker is recorded.
• Private routes use appropriate metadata and response headers.
• Third-party scripts load only when needed.

Deployment

• The recorded Go for Launch revision is still current with its configured upstream, or a reviewed pin is documented.
• The production candidate was built before staging and Simulator testing.
• The exact tested candidate is deployed to staging.
• Staging returns the candidate identifier expected by the release gate.
• PageSpeed audited the expected staging candidate rather than an error, stale, or access-denied page.
• All eight required PageSpeed category checks equal 100.
• No production push occurs before the Simulator and PageSpeed gates pass.
• Production deployment completes successfully.
• The canonical public hostname serves the new candidate.
• The production /sitemap.xml, child sitemaps, and robots declaration pass public HTTP checks.
• Public robots.txt returns HTTP 200 and advertises the exact canonical sitemap URL.
• A fresh approved external crawl was run after deployment, or the access blocker and required follow-up were recorded.
• Ahrefs API v3 Site Audit was checked when approved access exists, and the report records pass, fail, or an allowed skipped state.
• Apex and alternate-host redirects behave correctly.
• Opposite trailing-slash forms return one HTTP 301 or 308 to the exact canonical URL.
• Redirect probes preserve paths and query strings unless an approved map intentionally changes the path.
• Localized canonical, hreflang, language selector, and sitemap behavior passes on public hosts.
• Live WebKit tests pass after propagation.
• Native iOS Safari navigation works on production.
• Cloudflare edge HTTP errors were queried immediately after production when approved zone analytics access exists.
• Cloudflare RUM was compared after sufficient production traffic, with the report distinguishing available, no-data, skipped, permission-error, and failed states.
• Any required Cloudflare threshold or baseline regression passes under the reviewed enforcement mode.
• Generated build output is cleaned without reverting source changes.
• Final evidence records versions, test counts, scores, and remaining risks.
• Final evidence records the render sharpness result, report path, native-resolution review, and intentional exceptions.
• Final evidence records sitemap counts and Search Console verification and submission status.
• Final evidence records the AEO query baseline, reviewed pages, sources, and measurement plan when answer-focused content changed.
• Final evidence records the design mode, framework, status, findings, and evidence paths.