authoring-architecture-doc

Guides authoring of whole-system architecture documents with linked standalone ADR files.

1|1|Updated May 24, 2026
One-click install
npx skills add https://github.com/bm629/agent-skills --skill authoring-architecture-doc-bm629
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: authoring-architecture-doc
Source: https://github.com/bm629/agent-skills/tree/main/skills/authoring-architecture-doc
Command: npx skills add https://github.com/bm629/agent-skills --skill authoring-architecture-doc-bm629

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Writing a whole-system architecture document from an approved PRD often produces boilerplate outlines, unjustified technology picks, and decisions buried inline. This Skill supplies the method for producing a grounded architecture document: scoping the boundary first, naming one responsibility per component, justifying each significant technology choice, realizing every NFR target with a mechanism, and recording each key decision as a standalone, linked ADR file. ## Core Features & Use Cases - Method-driven authoring: Fills an architecture-doc template's sections with judgment — boundary and stakeholder concerns first, protocol per topology arrow, failure semantics per integration boundary, and diagram-narrative sync. - Standalone ADR mechanism: Records each significant decision as its own immutable ADR file (one decision per file, with driver and real alternative), keeping only a linked decisions index in the document; changes supersede rather than rewrite. - NFR realization and amend path: Gives every PRD quality target a documented mechanism (measurable quality-attribute scenarios where load-bearing) and supports delta-based amendment with supersede conventions and downstream ripple analysis to dependent technical-design docs. - Use Case: Given an approved PRD for a new product, produce the system architecture document — context, components, topology, tech-choice rationale, NFR realizations — plus a set of linked ADR files that a reviewing skill can gate against a 10-condition bar. ## Quick Start Use authoring-architecture-doc to write the system architecture document from the approved PRD in docs/specs/auth-v2.md, with each key decision recorded as a linked ADR file.

Frequently Asked Questions about authoring-architecture-doc

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

FAQPage Schema
How do I write a software architecture document from a PRD?

Start from the approved PRD and product direction, take the section structure from an architecture-doc template tool, then fill it with the method: boundary first, one responsibility per component, protocol per topology arrow, rationale per significant tech choice, and a realization mechanism per NFR target.

How should architecture decisions be recorded in a design doc?

Record each significant decision as a standalone ADR file (Status, Context, Decision, Alternatives, Consequences), one decision per file, and keep only a linked decisions index in the architecture document. Accepted ADRs are immutable; changes are made by writing a new ADR that supersedes the old one.

What is the difference between an architecture document and a technical design doc?

An architecture document covers the whole system: components, topology, boundaries, technology choices, and NFR realization. A technical design doc covers one feature's implementation at a lower altitude and locates itself within the architecture document.

When should I not use an architecture document skill?

Do not use it to review a finished architecture doc (use the dedicated reviewing skill), to design a single feature's implementation (use a technical-design skill), or to specify every API endpoint or database table (use API-spec or data-model skills).

How do I amend an approved architecture document without rewriting it?

Scope the change to the affected decision, component, boundary, or NFR realization and edit in place. Change decisions only via a new superseding ADR, keep the decisions index in sync, bump the document version and changelog, and flag the ripple to dependent technical-design docs.