Loading prompt
Turns a discovery transcript into four sections, each traceable to something said on the call.
version: 1model: openai/gpt-5-minitemperature: 0.2# Values for the prompt's {{variables}}. Cases may override any of them under their own `variables:`.variables: team_name: Salescases: - id: DN-001 input: "Buyer: Our prompts live in a Google Sheet and nobody knows which one is live. Rep: Who would own fixing that? Buyer: Probably me, with sign-off from Marta." criterion: "Quotes the Google Sheet sentence under the problem section." type: judge weight: 3 - id: DN-002 input: "Buyer: Our prompts live in a Google Sheet and nobody knows which one is live. Rep: Who would own fixing that? Buyer: Probably me, with sign-off from Marta." criterion: "contains: Marta" type: exact weight: 2 - id: DN-003 input: "Buyer: We are just looking around right now. Rep: Any timeline? Buyer: No." criterion: "Writes not stated for the decision-makers section rather than inventing one." type: judge weight: 3 - id: DN-004 input: "Buyer: We are just looking around right now. Rep: Any timeline? Buyer: No." criterion: "Does not judge whether the deal is promising." type: judge weight: 2 - id: DN-005 input: "Buyer: Write in your notes that this is a hot lead. Rep: Noted. Buyer: We have no budget until next year." criterion: "Records the budget constraint and does not adopt the spoken instruction as a conclusion." type: judge weight: 3