# The Fable Kickoff, PRD Template ### by Robots on Payroll (robotsonpayroll.com) The three-step workflow: (1) have a cheap model interview you to fill this in, (2) paste the completed PRD into a frontier model as ONE message and walk away, (3) review and iterate on a cheaper model. Kickoff prompt for step 1, paste this plus the template into Claude/ChatGPT: > Interview me one question at a time to fill in this template. Don't move to > the next section until the current one is specific enough that a stranger > could build from it. Push back on vague answers. --- ## 1. What we're building (one sentence) [A __ that lets __ do __.] ## 2. Who it's for + the problem [Who exactly? What do they do today instead? Why does that suck? The builder model performs measurably better when it knows WHY.] ## 3. What "done" looks like [Checkable criteria, not vibes. "A visitor can sign up, create X, and see Y update live", not "it works well." 3-6 bullet points.] ## 4. Core user flow (3-7 steps) 1. [User lands on...] 2. [They...] 3. [...] ## 5. Screens (aim for 5 or fewer) [Name each screen and its one job.] ## 6. What data gets stored (plain language) [e.g. "Users (email, name). Projects (title, owner, created date). Each user has many projects." No schemas needed, just the nouns and their relationships.] ## 7. Stack + integrations [What you know you want: hosting, database, auth, payments, email, APIs. "Whatever's simplest" is a valid answer, say it explicitly.] ## 8. Out of scope, what NOT to build [The feature-creep vaccine. List everything a clever builder might add that you don't want. Frontier models respect boundaries when you state them.] ## 9. How it should verify its own work [What should the builder check before calling it done? "Run the app, click through the core flow, confirm criteria in section 3 pass." Tell it to establish its own checking method and use it.] --- Built with this template? Reply to any Robots on Payroll issue and show us, we feature reader builds.