# Content stewardship working packet

Six distinct reader tasks: collection inventory, finding/navigation structure, external-source evaluation, cross-format adaptation, structured PDF reading review and operation-level automation review. Each has complete bilingual contextual examples and a blank private record. All supplied inventory entries, maps, claims, files, observations and account IDs are constructed. No real participant, source investigation, PDF edit, native account operation, publication or business result is represented.

The example files reproduce the complete guide context, not only an outline. The reuse guide includes full text, document-page and video-script drafts. These are scripts, not generated PDF/video files or a recording of LinkedIn. Two original SVG diagrams explain claim dependencies; render-diagrams.mjs reproduces them from stewardship-fixtures.json using only existing Node built-ins. The diagram does not prove that every real copy will be found or that an export is accessible.

Run node check-examples.mjs in this extracted folder with an existing Node installation. It checks declared inventory counts, a six-card grouping, repeated and cyclic source lineage, fictional fractions, a four-page label mapping, one declared external ID observed twice and actual SVG markers. These checks are not semantic uniqueness, source truth, rights, WCAG/PDF-UA, API permission or reader-study certification. The examples make no real network request and perform no external action.

Primary documents were read on 1 November 2026. Government content policies are adapted with their original system boundaries; Govsearch does not inventory every blog. Card sorting can inform a proposal and does not supply the one correct architecture. The DfE history reports its own nine-participant work and later tree testing as a plan, not our study. Only the public Digital Inquiry Group introduction was read, not gated lessons or underlying papers. Commons Library was read through the official web body after direct403. The former Full Fact process URL returned410 and is excluded.

W3C PDF techniques concern specific requirements and are informative, not a full certification. Bookmarks, tags, title, language, page labels and actual reading are different checks. Default language does not translate the document. A descriptive transcript is not automatically a replacement for required AA audio description. LinkedIn terms and developer authorization do not certify any vendor or grant every action; this packet supplies no evasion or account-safe action limit.

Keep completed private records minimal and outside public copies. The identity manifest excludes itself and the ZIP to avoid circular hashes. The archive includes every source file plus that manifest. No paid collection, outreach, recruitment, off-site posting, new schedule or independent reviewer was used.