Books, overviews and speaker cards — evidence

Bounded preparation, 5 October 2026. This checks three parts of B-06. It does not certify the full taxonomy, current assignments or the deployed data path. No new meaning is approved.

History recovered

The original B-06 card in docs/atlas/_data/review/content-identity.json bundles eBook, One-Pager, Speaker Card, Email Header, Blog Header Image, Social Post Graphic and Wallpaper. Its recommendation to exclude all seven is a proposal, not Joe's answer. The card's claim that none exists anywhere is not a current system fact.

Read the current Content chapter and Brand & Identity review, the relevant approved-type list, Decided's ratification section and the original August reply in docs/reference/2026-07/2026-07-26/atlas-review-responses.md. Template belongs to the approved type list; B-06 does not need to reopen its existence. The August 16 source request in docs/reference/2026-08/2026-08-16/brand-identity-retail-round-notes.md authorizes selective extraction and review, not blanket adoption of source labels.

The existing rules for whole documents, useful parts, versions, publications, captures and photographs apply. Focused searches did not locate individual approvals for these seven additions or a Book type. That is a bounded recovery result, not proof that every prior conversation was accessible. Book is therefore explicitly proposed in the new question, rather than passed off as an existing approval.

Inspected examples

• Supplied archive/reference/brand-guidelines-ref/github/GitHub Brand Guidelines 2025.pdf: rendered and viewed complete pages 61–63. Text extraction returned formfeeds, so it was not used as evidence of absence. Page 61 distinguishes libraries and an asset-generation tool. Page 62 shows a Speaker Card template preview with portrait, talk title and branding. Page 63 lists ready-to-use deliverables including eBook Covers, and distinguishes assets from complete key art. The tool preview does not prove publication of the finished card. No private Drive or Figma files were accessed.
• Kiva/OpenAI ChatGPT Basics PDF: both pages rendered and inspected. The filename says One-Pager; the file contains two pages. It is a concise introductory handout. This supports a question about a functional name versus literal extent, not automatic classification from the filename.
• Opensource.com collection page and its linked PDF: the page offers PDF and EPUB. The PDF's cover, contents and introduction were rendered and inspected. The source calls it both a collection and a booklet. The proposed Book boundary includes collected volumes; it is not asserted as an industry-mandated classification. Remaining pages were not visually audited. No comparison of the EPUB's contents was performed.
• W3C EPUB 3.3, abstract: EPUB specifies packaging and encoding for digital publications. That supports separating file encoding from semantic form. It does not prescribe BrandTrackers' type names or prove equivalence between two files.

Source hashes and inspection limits are retained in books-overviews-source-observations.json. The new examples are external references, not claimed library records. B-06's two older image records were queried, but their images were not inspected in this unit and they do not establish new classifications.

Selected system checks

books-overviews-live-evidence.json preserves literal read-only SQL and results. Schema probes preceded references to the selected columns. An exact-name query of schema_enums returned Template under element, with no description. It did not establish implementation of the approved whole-piece Template type. The first broad regex also matched Facebook accidentally; that false positive is excluded and the exact-name query controls this finding. Absence from a selected-name result is not absence from every store or alias system.

The two old-card records returned as image format with blog_post and social_post content types respectively. Those values alone cannot establish the intended graphic form. Their causes and full evidence relationships remain unverified; no labels were changed.

Selected local consumers were inspected in backend/services/post_processor.py: content_type is among scalar updates, while elements use a separately loaded vocabulary. frontend/src/lib/build-content-query.ts filters content_type for content-type searches. This is a limited local trace. A new Docs type is not proof that prompts emit it, the validator accepts it, the writer preserves it or all consumers can browse it. No deployed writer or full document capture path was tested here.

Independent architecture assessment

Keeping the older blanket exclusion is simple but loses useful distinctions between a complete publication, a short overview and a designed speaker introduction. Adding every source label as a type is also weak: source folders combine reusable inputs, outputs and placements. Selective additions with explicit boundaries preserve useful research paths without converting every filename into a canonical value.

Book plus a digital-edition browse path avoids making encoding define the work. A separate eBook type remains a reasonable alternative if that is the preferred human vocabulary. One-Pager could be a whole-piece type or a presentation description; the packet makes that choice explicit. Speaker Card could similarly be a type or a broader graphic with a defined purpose. None of these choices approves an exact physical schema or claims these boundaries are universally standardized.

For the comprehensive audit, trace source URL and document bytes → page extraction → parent/part identities → classification inputs and composed prompts → validation → writes → publications → search, trackers and citations. Compare actual editions before deduplicating. Preserve both publisher wording and observed page count. Inspect relationships between reusable templates and finished outputs; a preview is not proof of derivation. Distinguish intended use from an observed publication.

The implementation contract must inspect keys, required values, multiplicity, junction uniqueness and indexes against real queries. Include multi-author books, repeated publication, multi-speaker cards, uncertain event membership, corrected identities and partial extraction. Permissions must apply to parent documents and independently browsable parts. Reruns and monitoring must preserve human corrections, source revisions and prior citations; they must not infer that a new export is a new campaign. Scheduling, costs and delivery remain under the existing activation gates.

Queue and completion

B-06 is retained as the original seven-part obligation. Its Book/eBook, One-Pager and Speaker Card parts are prepared separately. Email Header, Blog Header Image, Social Post Graphic and Wallpaper remain open individually. No answer to one closes the others. The current identity proposal should link to this clarification while preserving the original card unchanged.

Readiness requires independent challenge, checks, rendering and hosted verification. The complete Docs reconciliation and database/pipeline audit remain outstanding. No schema, production data, prompt, API or automation setting changed.