{
  "site": "https://www.nhs.uk",
  "packetSHA256": "00391b37631387be8a4df161c5b4b6877e86c48588a4dbac55885b30fb784eda",
  "reviewer": {
    "harness": "Codex native subagent",
    "model": "gpt-6.1-sol",
    "reasoning": "high"
  },
  "rows": [
    {
      "checkId": "distraction",
      "guideline": "2.5",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6",
        "probes.json:journeys:0"
      ],
      "rationale": "Across all seven sampled page states, task headings and navigation remain available. The cookie notice offers a relevant privacy choice and is dismissed during the walking task; the exercise page presents paused, user-started instructional videos. Hidden player dialogs and empty search status regions are not observed interruptions. This pass covers the captured reading states only.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "choice",
      "guideline": "2.6",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6",
        "staged.json:runs:0",
        "staged.json:runs:1"
      ],
      "rationale": "Accept and reject are plainly worded, adjacent similarly sized buttons, and both branches complete. Optional cookie settings are initially off. Geometry and text do not establish visual prominence, contrast, or default focus; no screenshot/style comparison establishes parity across the sample. No undue steering is demonstrated.",
      "confidence": "medium",
      "nextEvidence": "Capture banner appearance and default-focus behavior for both paths, including narrow presentation."
    },
    {
      "checkId": "comprehension",
      "guideline": "2.7",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "All seven pages have descriptive task headings, familiar navigation labels and readable explanations. Walking uses question headings and concrete examples; videos have named exercises, short summaries and discoverable safety information. Cookie settings explain optional categories and purposes. This is an English reading judgment, without a claim of representative-user comprehension or clinical accuracy.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "consistency",
      "guideline": "2.8",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "The seven pages repeat the same primary navigation names, NHS home link and labeled search. Content pages consistently add breadcrumbs and topic/policy navigation. This supports interface naming consistency in the sample; it does not prove design-system documentation.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "motion",
      "guideline": "2.10",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Screen and reduced-motion snapshots show no running animations, and all 15 instructional videos are paused without autoplay. However the video print state records 15 running one-iteration DIV animations after preference changes; target purpose, duration and visibility are not established. Playback controls are present but activation was outside the task. No harmful or uncontrolled motion is demonstrated.",
      "confidence": "medium",
      "nextEvidence": "Inspect the recorded DIV animation effects and visible behavior under reduced motion, and activate one representative player to confirm playback controls."
    },
    {
      "checkId": "notifications",
      "guideline": "2.12",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "No Notification.requestPermission calls occur in any captured state. The cookie notice is a useful choice; video player live regions are control/status feedback, with hidden dialogs rather than observed unsolicited alerts. Browser permission was already denied, so this supports only the observed reading states, not future notification behavior.",
      "confidence": "medium",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "documents",
      "guideline": "2.13",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "No document downloads are listed on the seven sampled pages. Print emulation retains task headings and reading summaries while removing primary navigation; Walking retains article text and the video page retains exercise summaries and posters. No PDF optimization, paper page count or complete video transcript equivalence is claimed.",
      "confidence": "medium",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "unused-code",
      "guideline": "3.2",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Initial coverage identifies unexecuted ranges in shared consent and main scripts and the video player. Coverage lasts only initial load plus 500 ms, has truncated function lists, and does not cover search, account, settings updates or playback. Consent and navigation execute later, so no code removal is justified from these offsets.",
      "confidence": "medium",
      "nextEvidence": "Collect representative interaction coverage and identify concrete unused modules before proposing removal."
    },
    {
      "checkId": "components",
      "guideline": "3.3",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:4",
        "staged.json:runs:8",
        "staged.json:runs:9"
      ],
      "rationale": "All 15 paused video players initialize before playback, with a 316,598-byte compressed player script, per-video metadata and HLS segment requests. The staged initial video run records 21,825,011 received bytes with partial accounting before consent or playback. This is a strong deferral investigation target, but the packet does not exercise a facade or establish usable on-demand activation; do not claim a proven split or savings. Shared scripts on the remaining six pages have no corroborated removable component.",
      "confidence": "medium",
      "nextEvidence": "Prototype one player initialized on request and verify transcripts, safety text, captions and playback remain usable; compare a bounded repeat measurement."
    },
    {
      "checkId": "layout",
      "guideline": "3.10",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Every page retains its task heading and core text at desktop and 360 CSS-pixel width, without recorded horizontal overflow. Narrow layouts change primary navigation into a Browse/More control, but the collection does not activate that affected control. Width emulation alone cannot establish that route works.",
      "confidence": "medium",
      "nextEvidence": "Open and use the narrow Browse/More navigation on the sampled templates and confirm the intended task link can be reached."
    },
    {
      "checkId": "impact-choice",
      "guideline": "5.4",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:4",
        "staged.json:runs:8"
      ],
      "rationale": "Routine reading and cookie controls do not need carbon labels. Choosing an exercise video could justify a data-use or text-alternative cue, but every player is already initialized before play and the transcript controls are not exercised. The evidence does not establish a working lower-data choice or an appropriate supported estimate.",
      "confidence": "medium",
      "nextEvidence": "Exercise a transcript and player selection; establish useful data-use guidance from measured behavior before adding it."
    },
    {
      "checkId": "journeys",
      "guideline": "2.4",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:journeys:0",
        "context.json:records:0"
      ],
      "rationale": "The declared pointer journey rejects analytics, follows Healthy living then Exercise then Walking for health, and reaches the correct article URL, heading and full body. Completion is supported by the captured content and independent task trace, not successSelectorVisible (which is null). This does not establish optimal length or other audience routes.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "media-delivery",
      "guideline": "2.9",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "The DOM image inventory contains no task images on six pages, but it omits CSS backgrounds: the homepage network retrieves a roughly 475 kB PNG (test5_4nG3bKJ.width-1000.png), whose visibility and task value are not established. The video page has useful near-1280 by 720 posters at about 629 by 354 desktop and 328 by 185 narrow without srcset. These are investigation candidates; desktop density can justify the poster dimensions, and neither a smaller legible rendition nor removal of the background has been validated.",
      "confidence": "medium",
      "nextEvidence": "Inspect poster detail and alternative renditions at intended device densities before changing image delivery."
    },
    {
      "checkId": "font-delivery",
      "guideline": "2.11",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "The seven English pages use two WOFF2 Frutiger faces, regular and semibold, with display:fallback and Arial/sans-serif fallbacks. Observed faces serve body text and headings; the two font responses are about 17 kB each with validators and cache lifetimes. This meets the scoped strategy expectation without claiming constrained-loading validation or recommending unverified subsetting.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "semantic-forms",
      "guideline": "3.6",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "All seven initial documents have lang=en, one meaningful H1 and one main region. Search has an associated Search the NHS website label and a named submit button; cookie radios have explicit labels; visible video Play and Transcript controls have text/title names. Hidden caption controls with aria-labelledby are not alleged to lack names. This is the bounded document/control check, not full semantic validity or WCAG compliance.",
      "confidence": "medium",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "metadata",
      "guideline": "3.8",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Each sampled page has a descriptive title, matching absolute canonical URL and purpose-specific description. Structured-data blocks, where present, parse as JSON; the homepage declares none. Schema correctness and search eligibility are outside this check.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "preference-behavior",
      "guideline": "3.9",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Live and cold dark/reduced-motion observations retain task text and use the same light NHS palette. A light design alone does not establish an applicable unmet theme requirement. No screen motion is recorded, but the video print state records running DIV animations while reducedMotion is true, without enough effect information to judge purpose or response. Preference loss or a theme defect is not demonstrated.",
      "confidence": "medium",
      "nextEvidence": "Inspect the video DIV animation effects under reduced motion; establish any applicable theme behavior from design intent or a user need before prescribing dark mode."
    },
    {
      "checkId": "no-js-content",
      "guideline": "2.17",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Fresh no-JavaScript observations on all seven pages preserve meaningful headings, reading text and navigation. Walking remains complete; video headings, summaries, review dates and safety links remain available for the declared task of finding information without playback. My account, enhanced player controls and settings submission are outside this core-reading check.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "api-requests",
      "guideline": "3.11",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6",
        "staged.json:runs:8"
      ],
      "rationale": "Ordinary pages make a small emergency-banner request whose freshness has a plausible safety purpose. The video page requests metadata for 15 distinct videos and multiple HLS resources before playback, with many tracker events. Distinct video IDs and tracker event types are not redundant polling merely because endpoint paths repeat. The bounded capture does not establish recurrence needs, backend cost or safe removal.",
      "confidence": "medium",
      "nextEvidence": "Trace one player initialization and playback with initiators and timing, separating required metadata/captions from avoidable pre-play segment or analytics requests."
    },
    {
      "checkId": "public-cache-policy",
      "guideline": "4.2",
      "outcome": "pass",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "Sampled reusable NHS CSS/JS, fonts and poster/media responses expose cache lifetimes and/or ETag/Last-Modified validators. Fingerprinted CSS/JS have long public lifetimes; cookie-consent has a day lifetime; myaccount.js has validators. Brightcove metadata no-store and tracker no-cache responses have different purposes and are not failed as static assets. Header policy supports this narrow pass; cache reuse and offline support were not measured.",
      "confidence": "medium",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "text-compression",
      "guideline": "4.3",
      "outcome": "fail",
      "evidenceIds": [
        "probes.json:pages:4",
        "staged.json:runs:8"
      ],
      "rationale": "NHS HTML/CSS/JS and the Brightcove player script use gzip or Brotli. However 15 successful Brightcove playback JSON responses of about 8.6 kB each have no content-encoding, corroborated in staged resources. These are material text responses, not tiny empty bodies, errors or compressed media. This is a bounded third-party text-delivery contradiction; achievable byte or energy savings are not measured.",
      "confidence": "high",
      "nextEvidence": "Confirm content negotiation and enable beneficial compression for playback JSON through the media provider; retain already-compressed media as delivered."
    },
    {
      "checkId": "keyboard-path",
      "guideline": "3.6",
      "outcome": "cantTell",
      "evidenceIds": [
        "probes.json:pages:0",
        "probes.json:pages:1",
        "probes.json:pages:2",
        "probes.json:pages:3",
        "probes.json:pages:4",
        "probes.json:pages:5",
        "probes.json:pages:6"
      ],
      "rationale": "The 12-step traces follow visible consent, skip, logo, search, account and primary navigation controls in a meaningful order, with focusVisible true. Every recorded outline is transparent; borders, backgrounds and shadows were not captured. Neither invisible focus nor a discernible indicator is established. The public accessibility statement reports focus contrast issues but does not identify these exact states.",
      "confidence": "medium",
      "nextEvidence": "Capture focus screenshots or computed border/background/shadow changes for representative links, buttons and search fields."
    },
    {
      "checkId": "disclosure-scope",
      "guideline": "2.1",
      "outcome": "cantTell",
      "evidenceIds": [
        "staged.json:discovery:0",
        "disclosures.json:sites:0"
      ],
      "rationale": "The bounded public discovery retrieved 404s for two common sustainability paths and no usable service-specific statement. That is insufficient to establish absence of an NHS website sustainability disclosure elsewhere; no retrieved statement can be judged for audited-service scope.",
      "confidence": "medium",
      "nextEvidence": "Supply a public statement explicitly identifying the NHS website and its impacts and mitigation."
    },
    {
      "checkId": "carbon-discovery",
      "guideline": "3.13",
      "outcome": "fail",
      "evidenceIds": [
        "staged.json:discovery:0"
      ],
      "rationale": "Root and well-known carbon.txt routes return 404; the homepage supplies no Carbon-Location and DNS lookup yields NoAnswer. No parseable document, redirect/delegation or version can be evaluated in this declared discovery scope. This is the narrow carbon.txt absence only, not evidence of high emissions or failure of the whole guideline.",
      "confidence": "high",
      "nextEvidence": "If useful to the service, publish a valid discoverable carbon.txt with relationships backed by evidence; do not treat it as an emissions reduction."
    },
    {
      "checkId": "hosting-evidence",
      "guideline": "4.1",
      "outcome": "cantTell",
      "evidenceIds": [
        "staged.json:discovery:0",
        "disclosures.json:sites:0"
      ],
      "rationale": "The frozen discovery contains no provider/service sustainability evidence linked to this origin. CDN and asset hostnames cannot establish operational electricity sourcing, provider boundaries or sustainable hosting.",
      "confidence": "medium",
      "nextEvidence": "Provide a documented origin/provider relationship and independently corroborated operational evidence for that hosting service."
    },
    {
      "checkId": "access-barriers",
      "assessmentScope": "Anonymous English readers using a desktop browser with pointer input and JavaScript, 1440 by 1000 CSS pixels: start at https://www.nhs.uk/, reject analytics cookies, follow Healthy living > Exercise > Walking for health, and reach the article heading and full reading content. No account, symptom input, booking or submitted health data. Keyboard-only, assistive-technology and constrained-device completion are outside this assessment scope.",
      "guideline": "5.13",
      "outcome": "pass",
      "evidenceIds": [
        "context.json:records:0",
        "probes.json:journeys:0",
        "context.json:records:7"
      ],
      "rationale": "The complete 1440 by 1000 pointer/JavaScript trace rejects analytics and reaches the full Walking article without login, submitted health data or access gate. No material barrier blocks the exact declared anonymous English desktop task. The accessibility statement is attributed broader context and does not invalidate the successful pointer route or establish accessibility for excluded audiences.",
      "confidence": "high",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "data-minimization",
      "assessmentScope": "Anonymous visitor analytics-cookie choice on the public NHS website on 2026-11-06: inspect accept/reject and settings choices, analytics-data categories and stated purpose across the homepage banner, cookie details and public privacy notice. No health-data task, identity form submission, NHS login or NHS App processing. Retention enforcement and cross-service propagation cannot be established by this observation.",
      "guideline": "5.18",
      "outcome": "pass",
      "evidenceIds": [
        "context.json:records:4",
        "context.json:records:5",
        "context.json:records:6",
        "probes.json:pages:5",
        "staged.json:runs:0",
        "staged.json:runs:1"
      ],
      "rationale": "The anonymous analytics choice offers direct accept/reject plus settings, with analytics, pop-up and campaign categories initially off. Settings and cookie details explain page/link usage, visit/session identifiers, purposes and durations; privacy text relates analytics to improving the service and logs to availability/security. Reject leaves reading available and accept adds analytics resources in the observed homepage branch. This supports the declared choice/purpose scope; no retention enforcement or cross-service propagation is inferred.",
      "confidence": "medium",
      "nextEvidence": "No further evidence is needed for this bounded expectation; recheck after relevant changes."
    },
    {
      "checkId": "content-lifecycle",
      "assessmentScope": "NHS Healthy living exercise editorial collection, specifically Walking for health and Strength and Flex how-to videos, as publicly presented on 2026-11-06: assess the published review schedule, responsible editorial/clinical roles and evidence of an applied review/update/archive decision; retain historical material when it has continuing user value. No CMS history or backend deletion scope.",
      "guideline": "5.18",
      "outcome": "cantTell",
      "evidenceIds": [
        "context.json:records:1",
        "context.json:records:2",
        "context.json:records:3"
      ],
      "rationale": "The policy publishes at-least-three-year review and editorial/clinical responsibilities. Video page/media labels report 2025/2026 reviews, while Walking still says last reviewed 15 December 2022 and next due 15 December 2025 at capture. These dates flag a useful reconciliation need but do not establish the actual review decision, its rationale or an applied archive/update example. The supplied policy and date labels alone cannot pass execution or justify deleting useful exercise information.",
      "confidence": "medium",
      "nextEvidence": "Provide one dated review/retention/update decision for this exercise collection, resolve the Walking due-date discrepancy, and link a historical version only where continuing user value warrants it."
    }
  ],
  "topFindings": [
    {
      "evidenceIds": [
        "staged.json:runs:8",
        "staged.json:runs:9",
        "probes.json:pages:4"
      ],
      "summary": "Investigate all-player initialization and video-segment transfer before anyone selects an exercise.",
      "checkId": "components",
      "priority": "high",
      "title": "Investigate all-player initialization and video-segment transfer before anyone selects an exercise.",
      "recommendedAction": "Prototype initializing one chosen player on demand while keeping summaries, safety information and transcripts usable; compare transfer and activation. The observed 21.8 MB initial count has partial accounting and is not a savings estimate.",
      "status": "evidence gap"
    },
    {
      "checkId": "media-delivery",
      "priority": "medium",
      "title": "Investigate the homepage background image requested outside the DOM image inventory.",
      "evidenceIds": [
        "staged.json:runs:0",
        "probes.json:pages:0"
      ],
      "summary": "The homepage retrieves test5_4nG3bKJ.width-1000.png, recorded as 475,879 received bytes in the staged run, while the DOM image inventory is empty. CSS background visibility, purpose and necessary rendition quality are unresolved.",
      "recommendedAction": "Locate the background and inspect its visible role in the homepage task; remove or resize it only after confirming useful content, legibility and appropriate pixel density. No savings is established.",
      "status": "evidence gap"
    },
    {
      "evidenceIds": [
        "probes.json:pages:4",
        "staged.json:runs:8"
      ],
      "summary": "Fifteen material Brightcove JSON responses lack compression.",
      "checkId": "text-compression",
      "priority": "low",
      "title": "Fifteen material Brightcove JSON responses lack compression.",
      "recommendedAction": "Ask the media provider to verify negotiated compression for playback JSON and validate the result with response headers and transfer measurements.",
      "status": "observed defect"
    },
    {
      "evidenceIds": [
        "context.json:records:1",
        "context.json:records:2",
        "context.json:records:3"
      ],
      "summary": "Reconcile the overdue Walking review label and document an applied exercise-content decision.",
      "checkId": "content-lifecycle",
      "priority": "medium",
      "title": "Reconcile the overdue Walking review label and document an applied exercise-content decision.",
      "recommendedAction": "Confirm whether the review occurred, update its public status accurately, and retain useful reading content. Supply the dated decision rather than treating age as grounds for deletion.",
      "status": "evidence gap"
    }
  ]
}
