# Structure and language review ## Section-heading refinement: 9 October 2026 The owner found the boxes around standalone headings unnecessary. The implementation page made the problem clear: a boxed heading competed with the table header directly below it. All 20 standalone section headings now use plain heading text with more space above. This applies to the implementation, rules and inspection pages. Entry panels retain their title bands, which show where an individual record starts. Heading levels, wording and anchors remain unchanged. ## Presentation review: 9 October 2026 The owner requested clearer grouping across the wiki, including cards, sections and boxed headings. This pass reviewed the structure of all 23 Wiki pages. It inspected eight rendered views across seven main reading routes before changing the presentation. The earlier language review remains a separate assessment. ### What helped and what changed The Executive Summary, descriptive page titles, topic indexes and exact source links already support the main reading task. The main weakness was visual grouping. Repeated recommendations, duties and questions had similar headings and labels, so one entry could appear to run into the next. | Reading route | Observed difficulty | Presentation decision | |---|---|---| | Home | Three reading routes were embedded in a list with uneven amounts of detail. | Use three short navigation cards. Keep the Executive Summary and scope as ordinary prose. | | Inspection findings | The annual account, primary reports and regulatory comparison needed stronger boundaries. | Add short routes at the top, section bands for major divisions and open panels for repeated findings. | | Recommendations | Consecutive recommendation entries looked alike. | Give each existing R entry a visible boundary and title band. | | Rules and responsibilities | Legal scope, duties and enforcement used several levels of similar headings. | Keep the heading hierarchy; add major section bands and separate duty entries. | | Internal policy questions | Long lists could obscure the start and end of each CHECK entry. | Keep each CHECK in an open panel, with its full questions and evidence limits visible. Group the longer CHECK-13 list by responsibilities, staff suitability and candour. | | Implementation tracker | Repeated response and delivery labels could merge across recommendations. | Group each R entry; retain the comparison tables. | | Source register | The renderer reserved too much width for short IDs and too little for currentness notes. | Size only the six-column source tables for their content. Retain native rows, exact anchors and contained scrolling. | | Defence implications | The five themes needed clearer separation on a long page. | Use one open panel per theme; retain the wider lessons and limits below. | The remaining pages were checked for their purpose and structure. Evidence matrix, glossary and validation queue remain lookup tables. Research method and research log retain headings and lists. Architecture and the three knowledge-map pages retain their diagrams and map controls. The four LLM Wiki pages and two project pages remain maintainer references. These pages do not need cards simply to match the longer topic pages. ### Design basis and tools The review used the Product Design audit workflow, browser screenshots and the rendered accessibility tree. The implementation uses the existing design tokens, native Markdown headings and static HTML containers. It adds no card library or content-hiding interaction. The [NHS card guidance](https://service-manual.nhs.uk/design-system/components/card) supports short navigation groups and advises against using cards to highlight long-form text. The long entries here are therefore single-column reference panels, not navigation cards. [GOV.UK accordion guidance](https://design-system.service.gov.uk/components/accordion/) recommends trying headings and anchor links first and keeping essential content visible. The [W3C heading guidance](https://www.w3.org/WAI/tutorials/page-structure/headings/) supports the retained logical heading levels. These are design references, not topic evidence or a claim of design-system certification. ### Verification and limits Before-and-after screenshots are retained as local verification artifacts under `output/verification/presentation-audit/`. They record eight desktop reading views and representative mobile views. The final verification receipt records the automated build, browser and link checks. Source-row links, R and CHECK anchors, both appearances, mobile containment and print rules require verification before publication. This is an editorial and interface review, not a reader usability study or a full accessibility audit. Visual grouping reduces the effort needed to distinguish entries; it does not remove legal complexity or establish how quickly another reader will find an answer. The source register still contains detailed provenance. Its tables can scroll within the page when the available width is narrow. ## Whole-wiki review: 9 October 2026 The review covered all 23 Wiki pages and eight underlying knowledge pages. It checked structure, language, terminology, links between topics and the separation of evidence from internal policy questions. It did not recheck external publications or reassess current law. The overall reading structure is coherent. Two corrections were needed: - The inspection reference linked three general reporting lessons to CHECK-07, which concerns neonatal reporting. Those links now lead to the relevant records and harm-escalation questions. CHECK-07 retains its neonatal scope. - The research method called 7 October the assessment date. It now identifies that date as the initial review and explains that later additions retain their own check dates. The 44 registered sources and 22 evidence claims reconcile. Recommendation numbering, the R12 date discrepancy and the distinction between proposed reforms, published duties and delivery evidence are consistent. Maintainer material remains separate from the subject reference. No broad structural or language rewrite is needed. ## Initial review: 7 October 2026 **Reviewed 7 October 2026.** The review covered Home, the topic pages, the evidence registers, the knowledge-map guides and the agent and maintainer pages. The reading test was whether a new reader could understand the main conclusion before entering detailed evidence. ### Main findings and changes | Reading step | Problem found | Change made | |---|---|---| | Arrive on Home | Routes and qualifications appeared before an answer. There was no Executive Summary. | Added a short assessment, five key points and three main reading routes. Kept the full scope and governing question below. | | Understand the Defence relevance | Abstract terms and repeated qualifications obscured the main message. | Rewrote the five lessons in direct language. Each gives the evidence, possible Defence implication and separate internal questions. | | Find a rule or inspection finding | The public comparison mixed legal scope, policy comparisons and observations in large tables. | Moved observations to an inspection page. Replaced paragraph-heavy legal and inspection tables with headed sections and clear labels. | | Find a specific recommendation or question | Readers had to know a number or scan long entries. Several questions shared one paragraph. | Added topic indexes to the recommendation and internal-question pages. Put each internal question in its own bullet. | | Check a source or unfamiliar term | Technical labels and maintainer routes competed with subject information. | Grouped sources and limits together. Put knowledge maps and maintainer guides under About this wiki. Added relevant glossary terms and clearer source-register orientation. | ### Evidence and checks The visual review used the rendered Home, Defence implications, comparison and source-register pages. Before-and-after screenshots are local verification artifacts. They are not topic evidence. Source-register tables remain detailed because they are lookup records; readers can reach each record directly from a source identity. The rewrite preserves the 17 recommendation entries, CHECK-00 to CHECK-17, registered source identities and the exact governing question. The old CQC comparison anchor points to the new inspection reference. The legal table's qualifications remain in the headed sections. Public evidence, possible Defence implications and internal questions remain separate. The review also added the five recommendations from the Cyprus primary inspection report, registered as [S44](https://thirlwall-wiki.pages.dev/sources/index.md#s44). This is new topic evidence, distinct from the reading review. ### Remaining boundary This was an editorial and rendered-page review, not a reader usability study or a full accessibility audit. It does not establish how quickly another reader will find an answer. The source and evidence registers remain detailed reference material. Further topic additions should follow the same summary-first pattern and avoid duplicating the Home summary. The [project status](https://thirlwall-wiki.pages.dev/project/status/index.md) holds the current bounded content priority. Completed structural work is not a reason to repeat the same task.