LibraryZenithRequest a breakdown
Build DocsStandardAI-agent readyAI-agent ready

Documentation reference

AI Agent Brief

An AI Agent Brief gives an AI coding agent the context it needs to build responsibly: product intent, constraints, repository map, data model, design rules, feature scope, acceptance criteria, and verification commands.

> inspect repo
> read design rules
> implement feature
> run build
> report diff
Tests
Scope

Overview

What this document is

An AI Agent Brief gives an AI coding agent the context it needs to build responsibly: product intent, constraints, repository map, data model, design rules, feature scope, acceptance criteria, and verification commands.

AI agentCodexCursorPromptBuild handoffAcceptance criteria

Fit signals

When to use it

  • You plan to use an AI coding agent to implement product work.
  • The PRD is ready but needs translation into build instructions.
  • The project has design, data, or architecture constraints the agent must follow.
  • You want repeatable prompts instead of one-off vague instructions.

Audience and ownership

Who uses it

Founders, product teams, and engineers using Codex, Cursor, Claude Code, Lovable, v0, or other AI build tools.

FounderEngineerAI builderProduct managerQA

Before and after

Where it fits in the build chain

Idea Brief
BRD
PRD
Research
IA
Design System
Architecture
SOW
QA Plan
Launch

Gather before writing

Inputs required

  • Agent task prompt template
  • Repository context checklist
  • Acceptance criteria block
  • Verification command list

What each part should cover

Section-by-section guide

Context

Describe the context clearly enough that a founder, teammate, or AI agent can use it without needing hidden context.

Repo map

Describe the repo map clearly enough that a founder, teammate, or AI agent can use it without needing hidden context.

Scope

Describe the scope clearly enough that a founder, teammate, or AI agent can use it without needing hidden context.

Rules

Describe the rules clearly enough that a founder, teammate, or AI agent can use it without needing hidden context.

Acceptance

Make the standard observable so the team can review, approve, or reject the output without guessing.

Verification

Describe the verification clearly enough that a founder, teammate, or AI agent can use it without needing hidden context.

Practical workflow

How to create and use it

  1. 01Summarize PRD and technical architecture into build-specific context.
  2. 02Define exact files, conventions, constraints, and non-goals.
  3. 03Write task prompts with acceptance criteria and verification commands.
  4. 04Review AI output against product, design, and engineering expectations.

Copy into your workspace

Starter template

  1. 01Summarize product goal, feature scope, and non-goals.
  2. 02Add repo structure, coding conventions, and design rules.
  3. 03Specify data model, UI states, edge cases, and acceptance checks.
  4. 04Include exact commands for validation.

Make the draft usable

Weak vs strong examples

Agent task

Weak

Build the dashboard page.

Strong

Implement the dashboard route using existing card, table, and filter patterns; do not change auth or billing; verify with `npm run lint` and `npm run build`.

Context boundary

Weak

Make it like the design.

Strong

Match the existing `BuildTopNav` spacing, 8px radius card language, dark shell colors, and active green text treatment used on the library detail pages.

How to know it is useful

Quality checklist

  1. 01The agent knows what to build, where to build it, and what not to touch.
  2. 02Design and engineering conventions are explicit.
  3. 03Acceptance criteria and verification steps are copyable.
  4. 04The brief prevents vague generation and unrelated refactors.

Before you hand it off

Document readiness checklist

  • The ai agent brief has a named owner and approval path.
  • The business or product decision this document supports is explicit.
  • Inputs, assumptions, and open questions are separated from confirmed facts.
  • Scope boundaries and non-goals are clear enough to prevent silent expansion.
  • The document can be handed to a teammate, agency, or AI agent without hidden context.
  • The agent knows what to build, where to build it, and what not to touch.
  • Design and engineering conventions are explicit.
  • Acceptance criteria and verification steps are copyable.

Generate the first draft

AI prompt

Draft prompt

Create an AI Agent Brief for this product feature. Include product context, repository map, files to inspect, design system rules, data model, feature scope, non-goals, constraints, UI states, edge cases, acceptance criteria, test commands, and final review checklist.

Avoid these

Common mistakes

  • Giving an AI agent a feature idea without repository context.
  • No acceptance criteria or verification commands.
  • Missing design system and data model constraints.
  • Allowing the agent to infer business rules from thin context.

Use this after the draft

Handoff assets

  • Repo-aware implementation prompt
  • Feature-specific acceptance criteria
  • Verification checklist
  • Follow-up review prompt

Suggested build path

What to create before and after

Before

Idea Brief

Keep this connected so the build stays traceable.

Current

AI Agent Brief

Use this page to create the working draft.

After

PRD

Keep this connected so the build stays traceable.

Zenith CTA

Need this shaped for your business?

Zenith can turn rough notes into a founder-ready ai agent brief.

Bring your idea, current docs, users, constraints, and open questions. We will help make the next decision clear.

Request a breakdown