What is a spec packet?

A spec packet is a structured, build-ready specification handed to an AI coding agent such as Claude Code, Cursor or Codex. It bundles the problem, the evidence behind it, numbered requirements with acceptance criteria, constraints, non-goals and open questions, so the agent can build the right thing without guessing what was meant.

Updated

What is in a spec packet?

Typical contents of a spec packet
PartWhy the agent needs it
Problem and target usersSo it optimizes for the right outcome when choices come up
EvidenceSo humans reviewing the work can check why each requirement exists
User flowsSo the behavior is specified end to end, not screen by screen
Requirements with IDs and prioritiesSo work can be sequenced and referenced (REQ-001, REQ-002)
Acceptance criteriaSo the agent and reviewers know when each requirement is done
ConstraintsSo it respects the stack, performance limits and existing conventions
Non-goalsSo it does not build adjacent features nobody asked for
Open questions and assumptionsSo it stops and asks, or flags a guess, instead of silently inventing
Test plan and likely filesSo it knows how to verify the change and where to start

A free spec packet template lays these parts out in Markdown.

Spec packet vs PRD

A PRD aligns people and can leave gaps for a conversation to fill. A spec packet is meant to be executed: an AI coding agent will not ask what "fast" means, it will pick something. So a spec packet replaces adjectives with acceptance criteria, makes scope explicit through non-goals, and resolves or flags every open question.

Is a spec build-ready? Hard blockers

  • No clear problem statement or target user
  • No non-goals
  • A P0 or P1 requirement without acceptance criteria
  • No described user flow
  • An unresolved assumption with high impact if wrong
  • Requirements that have not been broken into implementable tickets

How does Praxiom AI produce spec packets?

Praxiom AI generates a Spec Packet from your research: numbered requirements (REQ-###) with priorities and acceptance criteria, assumptions (ASM-###) with their impact if wrong, and evidence (EV-###) linked to customer quotes. Each packet gets a readiness score and band, from not usable to build ready, with explicit blockers. From the packet, Praxiom generates one implementation ticket per requirement and a coding-agent prompt for Claude Code or Codex.

See specs for AI coding agents or sign up.

Frequently asked questions

Why not just paste a PRD into Claude Code or Cursor?

You can, but a PRD usually lacks testable acceptance criteria, explicit non-goals and resolved open questions. The agent fills those gaps with guesses, and the result drifts from what you meant. A spec packet closes the gaps first.

What is spec-driven development?

A way of working with AI coding agents in which a written spec is created before any code and becomes the source of truth the agent builds from. A spec packet is the input that makes spec-driven development work on product decisions, not only on technical tasks.

Who should write the spec packet?

The product manager owns the problem, evidence, scope and acceptance criteria. An engineer should add constraints, likely files and the test plan, and review it before the agent starts.

How detailed should acceptance criteria be?

Detailed enough that someone else could check them without asking you. The Given, When, Then format works well: "Given a user with no projects, when they open the dashboard, then they see the create-project prompt."

Run your product discovery on autopilot.

Bring your interviews, tickets and surveys. Get scored insights, a PRD and tickets.