PRD — Capture Intent Before Any Agent Runs
Phase 1 of 4 in the AI Driven Development Methodology. Before a GitHub Copilot agent writes a single line, you write down what you actually want built — in a form the agent can reason about.
What a PRD is in this methodology
A PRD (Product Requirements Document) is a short, structured statement of intent, scope, and constraints — written before any agent session begins. It is not a formal enterprise spec. It is the smallest amount of writing that lets an AI coding agent understand what "done" looks like without guessing.
A working PRD typically answers:
- What are we building, in one or two sentences?
- What is explicitly in scope — and what is explicitly out of scope?
- What constraints must the solution respect (stack, security, performance, existing conventions)?
- What does success look like — how will you know the agent's output is acceptable?
Why this comes first. Every downstream phase — PLAN, TASK, IMPLEMENTATION — depends on the PRD being right. An agent given a vague instruction will confidently fill the gaps with its own assumptions. A PRD removes the guessing.
What goes wrong without one
Without a written PRD, "AI-assisted development" tends to look like a chat window full of ad hoc prompts: useful in the moment, but with no record of what was actually agreed. Scope drifts silently. The next agent session — or the next teammate — has no way to reconstruct the original intent. This is the gap between using AI tools and doing AI Driven Development.
How this looks in practice
A PRD for AAM's methodology is a short markdown file, not a lengthy specification document. It travels with the project, gets referenced by the agent at the start of a session, and gets updated when scope genuinely changes — not rewritten from scratch every time.
Start with the free PRD template
The Workflow Starter Kit includes a ready-to-use PRD-template.md — part of a four-part kit
(PRD, PLAN, TASK, IMPLEMENTATION) that works in any language or framework.