bmad-create-architecture

The best way to document architecture patterns for AI agents is running a step-by-step discovery workflow that generates an Architecture Decision Document. This approach uses frontmatter state tracking and user confirmation gates to ensure downstream agents implement coherently without conflicts.

An architecture decision document workflow is a collaborative, step-by-step discovery process that translates a PRD and related inputs into structured architecture, patterns, and boundaries. It uses frontmatter tracking for workflow state and requires user confirmation gates before proceeding to ensure multi-agent consistency.

Yes, you can resume an architecture workflow without losing progress because the system uses structured frontmatter tracking for workflow state. It loads confirmed inputs and initializes the architecture workspace, allowing you to pause and resume the step-by-step discovery process seamlessly.

Yes, you need a PRD to start generating architecture documentation because the workflow translates your PRD and related inputs into architecture patterns and boundaries. The process requires mandatory initialization from step files and loads confirmed inputs to produce validation artifacts for downstream agents.

To prevent AI agent conflicts when implementing architecture, generate validation artifacts and decision documentation that establish clear boundaries and patterns. This step-by-step workflow produces agent-friendly content with frontmatter state tracking so multiple agents can implement the same project coherently.

You should not use a step-by-step architecture workflow when your project lacks a defined PRD or when you need rapid prototyping without multi-agent coordination. The workflow requires mandatory initialization from step files and user confirmation gates, making it unsuitable for unstructured or single-agent implementation scenarios.