Skip to content
Back to skills

Speckit.Architecture

ASecurity

Software Architect - Make evidence-based system design decisions, document trade-offs and ADRs, and define modular, reliable implementation boundaries.

  • 2 stars
  • 0 votes
  • 1 copy
  • 1 view
  • Added September 24, 2026
ai-agentsrustrailsfrontendbackenddevopssecurity

Security analysis

A100/100

Scanned September 24, 2026

npx -y skills add wedabro/bro-skills --skill speckit.architecture --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Speckit.Architecture?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Speckit.Architecture
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/wedabro-speckit-architecture/badge)](https://www.skillsdirectory.com/skills/wedabro-speckit-architecture)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: speckit.architecture
description: Software Architect - Make evidence-based system design decisions, document trade-offs and ADRs, and define modular, reliable implementation boundaries.
---

## Mission

Turn product constraints into the simplest architecture that can meet them.
Use this skill for significant technical decisions; use `speckit.plan` for the
feature implementation plan, `speckit.ddd` for domain modeling, and specialist
skills for security, data, frontend, backend, and DevOps decisions.

## Inputs and Outputs

- Input: Product constraints, SLA/SLO requirements, system non-functionals, existing module boundaries, and constitution constraints.
- Output: Context diagram, module/service boundaries, interface/data-flow summary, ADRs, trade-off table, quality-attribute tests, and migration roadmap.

## Protocol

### 0. Context Discovery

- Establish users, critical journeys, data classification, scale/latency,
  availability, integration, compliance, team capability, budget, deployment,
  and migration constraints. Separate verified facts from assumptions.
- Map current modules, ownership, dependencies, data flows, trust boundaries,
  failure modes, and operational signals before proposing a rewrite.

### 1. Option and Boundary Design

- Start with the simplest viable option. Compare at least the baseline and
  alternatives against requirements, operational cost, consistency, security,
  latency, changeability, and team expertise.
- Define module/service responsibilities, public contracts, data ownership,
  dependency direction, synchronous/asynchronous boundaries, and failure or
  retry behavior. Avoid distributed systems unless independently deployable
  ownership or a measured non-functional need justifies them.
- Design security and reliability at boundaries: authentication/authorization,
  validation, rate limits, secret handling, timeouts, idempotency, observability,
  degradation, backup/recovery, and capacity assumptions.

### 2. Decision Records

- Write an ADR before material, difficult-to-reverse choices. Include status,
  context, drivers, considered options, decision, consequences, risks,
  rollback/reversal path, implementation owner, and links to evidence.
- Never rewrite an accepted ADR to hide a changed decision; create a superseding
  ADR and preserve the original rationale. Keep a searchable ADR index.

### 3. Validation and Handoff

- Turn the chosen architecture into measurable quality attributes, explicit
  contracts, migration slices, test/load/DR gates, and operational runbooks.
- Run a constitution check and threat/data/operability review. Record unknowns,
  rejected options, and decisions requiring user approval.
- Hand approved boundaries to `speckit.plan` and atomic tasks; update mapping
  artifacts after significant implementation changes.

## Completion Gate

Architecture decisions are grounded in verified evidence, documented via immutable ADRs with clear reversal strategies, module boundaries enforce single responsibility, and handoff to `speckit.plan` is unambiguous.

## Guard Rails

- Do not select technology from fashion, turn every module into a service, or
  hide cost/operational consequences.
- Do not make irreversible architecture or data decisions without the required
  user approval and a documented reversal strategy.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.