API: steer the warm-transfer handoff briefing

Warm transfer — shipped 2026-08-17 — hands a live caller to a person along with a spoken brief the platform generates from the call so far. A flow can now steer that brief with two optional fields on the transfer_to_number node config:

FieldWhat it does
briefing_instructionsShapes how the generated brief is written — what to lead with, what to leave out, how long to be. Applied under the platform’s brief-writing rules, never in place of them, so the quoted transcript stays evidence no instruction inside it can escape. It is not variable-rendered: it is read by a model rather than spoken, so a {{var}} here is left literal on purpose. Max 2000 characters.
briefing_templateReplaces the generated brief with a fixed line, for a handoff that must state the same facts every time. {{var}} tokens resolve against the same call variables as hold_message, and the reserved {{system__summary}} resolves to the generated brief — so a template can pin the facts first and still hand over everything else the call turned up. A template that never names {{system__summary}} skips summarising entirely. Max 600 characters.

Both are warm-mode only and ignored in cold mode. Set neither and you get today’s platform-generated brief; existing flows do not change. {{system__summary}} lives in the reserved system__ namespace that no customer variable may use, so it can never shadow a variable a flow already owns.