Most questions you simply type — a name, a date (YYYY-MM-DD), a numbered choice. But some ask for a record or a
reference: an answer that is an existing row, not a spelling. For those the walk prints a numbered pick-list — a
#, the row's name, and its id — and you choose one. The client on the matter is picked from the Project's own people;
a country or jurisdiction is picked from Navigator's seeded reference data. Choosing a row stores that row's id in the
answer, so {{country__of_birth}} renders the one canonical Mexico every time, never a near-miss spelling of it.
To script the walk instead of typing it, pass a --select flag for each pick-list — its value is the number the walk
printed, or the row's id — and an --answer flag for each typed question, in the questionnaire's order:
navigator site intake answer <notation-id> \
--select person__client=2 \
--select country__of_birth=114 \
--answer 1990-04-12
Already have the answers on record — a recorded intake call, an email thread, a prior questionnaire? Hand the walk a
transcript and it pre-fills what it can:
navigator site intake answer <notation-id> --transcript <client-call.txt>
Navigator runs the transcript through its coverage engine, proposes an answer for every question the transcript covers,
and walks you through those as Enter-to-accept defaults, each labeled proposed from transcript. Nothing is
silently accepted: you confirm or correct every proposal, and the questions the transcript never touched still prompt as
normal. The attorney is still the actor — the transcript is only a faster first draft.