PRD-driven development: definition and examples

A PRD gives product work a source of truth. For human teams, it aligns intent. For AI coding agents, it gives the agent a structured target instead of an open-ended prompt.

Why it helps agent workflows

AI coding agents are more useful when the task has boundaries. A PRD can define the problem, user stories, acceptance criteria, constraints, and non-goals. That makes each implementation loop easier to review.

Example

Instead of telling an agent “improve onboarding,” a PRD-driven workflow defines the onboarding problem, target user, screens or commands involved, required behavior, and what must pass before the work is done.

Hal applies this idea to coding loops by converting requirements into stateful implementation iterations.