bmad-architecture

Create and validate architecture spine documents recording binding system design decisions.

Updated Mar 14, 2026
One-click install
npx skills add https://github.com/ArchaonHW/MingGoRTS --skill bmad-architecture-archaonhw
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-architecture
Source: https://github.com/ArchaonHW/MingGoRTS/tree/main/.agents/skills/bmad-architecture
Command: npx skills add https://github.com/ArchaonHW/MingGoRTS --skill bmad-architecture-archaonhw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Teams building a system in separate parts drift apart on the decisions that must stay consistent — boundaries, dependency direction, state mutation, shared data ownership. This Skill produces an architecture spine: a terse document that fixes only those invariants, so independently built units stay compatible. ## Core Features & Use Cases - Architecture spine authoring: Create, update, or validate an ARCHITECTURE-SPINE.md with stable AD-n decision blocks (Binds/Prevents/Rule), a named design paradigm, a pinned stack table, and a Deferred section, distilled from an append-only memlog. - Coaching or fast drafting: Work decisions out interactively through open-ended elicitation, or draft the whole spine fast with [ASSUMPTION] tags for review; brownfield runs ratify conventions found in the existing codebase. - Deterministic review gate: Run lint_spine.py to catch placeholders, duplicate AD IDs, missing fields, and unpinned versions, then dispatch parallel reviewer subagents against the spine. - Use Case: Point it at a spec package or an existing repository and ask for the architecture spine; it records each decision in a memlog, distills the spine, runs the reviewer gate, and hands off to spec or epic breakdown workflows. ## Quick Start Ask the assistant to create the architecture spine for your project from the current spec or codebase, choosing the coaching path when prompted.

Frequently Asked Questions about bmad-architecture

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create an architecture document from a spec?

Invoke the skill with your spec package or raw idea and choose the Coaching path for guided elicitation or the Fast path for a full draft with [ASSUMPTION] tags. Decisions are logged to a memlog and distilled into ARCHITECTURE-SPINE.md with stable AD-n identifiers.

What is an architecture spine versus a full architecture document?

A spine fixes only the invariants that keep independently built units consistent — paradigm, boundaries, dependency rules, state mutation, data ownership. Structural details like stack and source tree are treated as seed that the code owns once it exists, keeping the document terse.

Can it derive architecture decisions from an existing codebase?

Yes. In brownfield mode it reads the real code and persistent facts to ratify conventions already present rather than inventing new ones. Decisions already settled by reality are tagged [ADOPTED] in the spine.

How does the reviewer gate validate an architecture spine?

The gate runs lint_spine.py first to catch placeholders, duplicate or non-monotonic AD IDs, missing Binds/Prevents/Rule fields, and unpinned stack versions. Then parallel reviewer subagents apply the good-spine checklist and configured finalize_reviewers lenses against the spine.

Does it work without an interactive user?

Yes. Headless mode infers intent, altitude, and inputs from the first message payload, records inferences as assumptions, runs the full reviewer gate non-interactively, and returns a JSON result with status, artifact paths, and open questions.

When should I not use an architecture spine?

If the input is too thin to build on, run a spec workflow first so requirements exist to bind. The spine records decisions, not rationale or full prose documentation; discussion documents and fuller renderings are optional add-ons produced at finalize.