Verbal identity — evidence

Approval update, 5 October 2026. Joe accepted the three displayed verbal-identity boundaries and the separate three strategy boundaries. The recorded decisions supersede their pending status. The original bundle remains partial: other strategy entries and audio voice are not blanket-approved. The evidence below records bounded preparation, not a complete system audit. Calls to action and Offer have since received a meaning decision; their exact vocabulary and implementation remain audit work.

Prior work

Recovered the original B-03 card, the August 16 dictation, the source spreadsheet's verbal rows and the current identity review. The old card describes the verbal list as uncontroversial; that is an earlier recommendation, not a user answer. The source spreadsheet distinguishes voice, tone, messages, taglines, naming and terminology. Its extraction does not ratify those definitions.

The existing decisions already separate source-documented strategy from inferred analysis and permit verbal and strategic work within an identity program. Those capabilities are not asked again. The reviewed current Content guidance also distinguishes who recorded a claim from how it is supported. The proposed meanings must preserve both distinctions.

Primary sources inspected

• Mailchimp voice and tone: read the complete text guidance. Supports general voice versus contextual adaptation. No claim that every company uses those names or has immutable voice.
• NC State Think and Do guidance: read the text, including its distinction between mantra and optional tagline use. Supports retaining source terminology and application scope rather than automatic synonym mapping.
• Red Hat brand architecture: read the text, including product and component naming differences. Supports rules for naming offerings, not a universal database hierarchy.
• Mailchimp word list: read the complete list and usage sections. Supports preferred spelling and usage separately from product naming.

The examples are textual guidance; no motion or visual-treatment classification is claimed. Earlier example IDs in the old card were not inspected for this packet and are not used as evidence.

Selected system evidence

Literal read-only queries and results are preserved in verbal-identity-preflight.json. Column inspection preceded selected registry and guideline reads. A search for named columns cannot exclude equivalent JSON or other storage.

The selected registry query returns tone and writing-style entries. Two concrete value/display pairs are dramatic → Empathetic and empowering → Authoritative. Their history, intended aliases and deployed consumers need inspection before a correction can be proposed. No mapping was changed.

Selected guideline rows use voice_tone and have empty source-URL values. Their descriptions remain stored assertions. Entity names were not joined, and these rows have not been attributed to a particular writer. Missing source URLs here do not prove there is no evidence elsewhere.

The local onboarding prompt asks for combined Voice & Tone sections and metadata containing a primary tagline and brand-voice attributes. The local display adapter can derive synthetic sections from metadata, using statement or attribute-pills modes. Its synthetic publication flag is an in-memory display value; it does not establish a database publication write.

A local content prompt asks for extracted verbal or visual key messages. The inspected campaign digest and embedding code consume those messages. This means a change to their meaning can affect grouping and retrieval, not just a label on a profile. These excerpts establish local consumers, not active deployed paths, complete prompt composition, or successful execution.

Human relationships and architecture

An identity owner, a source publisher and a credited writer can be different parties. A guideline can define several verbal elements; several pieces can express one message; one piece can express several messages. A source-specific rule can apply to a product or campaign without becoming a permanent company rule.

Exact wording, paraphrase, interpretation, translation and usage context must remain distinguishable. Similar names do not establish product-family or ownership relationships. Similar slogans do not prove a shared campaign. A source's claim can be explicit without being independently true.

Keeping one prose description is simpler but weakens comparison. Treating every observed tone or sentence as a brand rule overstates the evidence. The recommendation defines distinct meanings while allowing shared presentation and ordinary piece-level observations.

What the complete verification must trace

Follow capture and source-part identity into extraction, validation, storage and publication. Check whether composed prompts preserve quotation versus summary, source scope, language and uncertainty. Inspect every writer and reader before changing existing message fields or guideline sections.

Verify permissions, evidence links, revisions and human corrections across reanalysis, search, campaign suggestions, trackers and Ask. Monitoring should detect changed guidance while preserving the earlier source. Scheduling and cost controls remain part of that trace; preparing these definitions activates nothing.

Physical keys, foreign keys, junctions, uniqueness, constraints, indexes and query plans remain for the complete audit. No new table or JSON shape is specified here. The packet can proceed because its semantic choices do not depend on pretending those implementation checks have passed.

Prepared 5 October 2026.

Earlier reconciliation, before the later verbal and strategy reply

The current review and its linked primary text examples were checked again after the morning answers. No answer was found for the three displayed verbal-identity choices in the current answer records; they remain pending. The earlier wording that calls to action were unresolved is superseded by the recorded approval. No new notification is needed.

The selected earlier system evidence remains dated evidence. This consistency pass did not rerun its database queries or verify the deployed onboarding and content paths. The exact distinction between source-stated messages and extracted summaries remains important: existing field names do not establish the proposed meaning. The current local detail adapter reads call-to-action and key-message descriptions; the showcase adapter also reads a value proposition. These are local consumers to include in the later compatibility audit, not proof of complete search or source evidence.

Later approval recorded

The exact reply is preserved in docs/reference/2026-10/2026-10-05/verbal-strategy-joe-response.txt, with six separate mappings in verbal-strategy-joe-answers.json. It approves the six displayed recommendations only. The earlier pending statements above describe the checkpoint before that reply. The original bundle remains partial, and no storage, prompt, schedule or runtime change follows from recording these answers.