Evidence for service headings and credits

Bounded review, 4 October 2026. Corrected 5 October: the reader packet asks two residual choices; an earlier moment-credit approval has been recovered. It is not an implementation or full-system audit.

Prior answers and original questions

Read the complete August 11 source records craft-services-and-products-feedback.md and services-rename-and-products-page-ruling.md under docs/reference/2026-08/2026-08-11/. Joe names Services at the top and points to the specific rulebook level as Craft. His example puts Animation, Video Production and Photography under Creative Production. The recorded ruling leaves the middle level open. The current decided page preserves that boundary.

Original content-agenda.json A-08 recommends Discipline but includes stale wording about Execution below the tree. The master creative-taxonomy specification describes Execution as cross-cutting rather than a fourth tier. The new packet asks only the residual middle name. Multiple craft parents and the already-approved separation of visible craft from credits remain intact.

Original A-09 asks about a per-beat crew-role vocabulary. The initial review checked Q16 craft parentage but missed the explicit July 21 g-6c approval in theratificationcnooslejuly21.md, read against its original card in ratification-console.html. Joe described Buck contributing animation to particular moments as well as social content. Root verified the exact note; the recovered source is banked in docs/reference/2026-10/2026-10-04/moment-credit-prior-approval.json.

The new allow-or-defer capability question was therefore redundant and is withdrawn. A-09 is PARTIAL: contribution scope is approved, while exact role names and the need for a distinct moment-role vocabulary remain unresolved. Neither the old card's storage assertions nor this approval proves current support. Joe's Buck scenario is an intended example, not a newly checked credit. The original sent-question receipt remains historical evidence of the error.

Original content-retail.json R-03 asks where Interactive Kiosk belongs and offers no preference. Its old example is not inspected evidence of a kiosk. The new packet substitutes publisher documentation. The approved type-list-heading rule does not prove multi-heading membership. Two entrances for this type are an explicit new recommendation; the alternative retains one home plus contextual discovery. Other heading assignments remain open.

Inspected examples and limits

• Read the publisher's McDonald's Għargħur visitor guide, especially PDF page 8 and the preceding ordering-options page. The text explicitly distinguishes kiosk, counter and mobile ordering. Browser access met a security-verification screen; no challenge was solved. The PDF text was available through the document reader. No visual screenshot inspection or live ordering test is claimed.
• Read Cooper Hewitt's digital visitor experience, explicitly dated 2015–2022. It describes museum touchscreen tables and distinguishes hardware, interactive software and preview-video responsibilities. This is historical source evidence, not a claim of present availability or an assertion that a table is a kiosk.
• Read Studio Dumbar's DevDay case study. Its prose supports project collaboration and multiple outputs. It does not identify an individual for a specific interval in that prose. No motion interval is classified or attributed from this inspection. The positive animator-breakdown case is explicitly hypothetical.

Selected live system evidence

Verbatim SELECT responses are banked in docs/reference/2026-10/2026-10-04/headings-and-credits-live-evidence.json. Schema was inspected before selecting values. The creative vocabulary has Creative Production at level 0, Animation at level 1 linked to it, and Character Animation at level 2 linked to Animation. These selected relationships support the naming example; they do not certify the complete tree, all parents or every definition.

The selected registry returns existing content-type families and credit roles. It did not return Interactive Kiosk in the exact selected type query. That is not global absence proof and does not settle the conceptual recommendation.

The selected content_credit columns include content, contributor, role, discipline, provenance and evidence fields, with no moment column in that table's returned schema. This is a scoped observation, not a claim that no alternative representation exists. Selected DevDay rows are imported canary records. Their presence and directly-stated labels do not prove automatic extraction or correctness of each asserted attribution.

The earlier credits-review-preflight.json retains both credit-table column probes. It is a separate historical receipt, not replaced by this one.

Local dependencies and defects to investigate

Read the credit-extraction block of backend/routers/pipeline_orchestrator.py. It reads credits from analysis results, resolves an entity by name and writes v6_content_credits. Its selected lookup and insert use content_item_id; the saved live schema returns content_id. The insert also supplies entity_type. This is a local code/schema mismatch lead. Deployed execution, all callers and failure handling still require tracing before calling it a reproduced production defect or designing a repair.

Read frontend/src/hooks/useEntityCredits.ts and useEntityCreditedContent.ts completely. Both read v6_content_credits; the latter uses credited content identifiers to fetch content. These local readers do not establish a working moment-credit journey or compatibility with the other credit table. The complete audit must identify which writer/read path owns canonical attribution before extending it.

The reference generator and vocabulary probes use creative_disciplines. Changing presentation terminology must not blindly rename table fields or identifiers. Inspect generated output, filters, composed prompts and stored saved-query rules as consumers in the later contract.

Human model and implementation contract to follow

Keep three questions distinct: what kind of work is visible, who contributed, and exactly what they contributed to. A depiction link is not a credit. A project credit is not automatically evidence for every asset, photograph or interval. Several contributors and roles can apply; unknown identities and scopes must remain representable without invented links.

Before implementation, inspect keys, foreign keys, permitted targets, uniqueness, relationship direction, permissions, indexes and actual query plans. Define how evidence survives revised boundaries, version changes, merges, source withdrawal and reanalysis. Distinguish removing an unsupported assertion from deleting its historical evidence. Establish correction precedence and migration compatibility for both credit paths.

Trace collection → source evidence → composed prompt → validation → write → publication → search and cited results. Scheduling can refresh source claims but must not upgrade uncertainty or erase human corrections. Heading changes must preserve type identity and equivalent browse/tracker rules. No processing run, schema mutation, prompt change or automation activation is authorized by preparing this packet.

Queue mapping

• A-08: middle level name only; Services and Craft stay settled.
• R-03: Interactive Kiosk heading placement and its proposed cross-listing only.
• A-09: contribution scope recovered from July; the redundant capability question is withdrawn. Exact role vocabulary remains open.

The original inventory and original cards remain intact. This packet does not close neighboring definitions, broader heading lists or the comprehensive audit.