Software Architecture

SkillMedia

Design and review software architectures from business drivers through system boundaries, tradeoffs, runtime behavior, evolution, and architecture practice. Use when choosing a greenfield or target architecture, comparing modular-monolith and service shapes, designing distributed consistency, replication, partitioning, coordination, ordering, transaction isolation, or failure behavior, defining architecture fitness evidence, or facilitating a consequential architecture review. Do not use for reverse engineering, API contract semantics, data-platform design, implementation, infrastructure operations, security lifecycle, or migration execution; route those to the named specialist skills.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Software Architecture skill

What this skill tells your AI

The instructions your AI receives, as published by magnus919/agent-skills in software-architecture/SKILL.md and read by ahel’s review.

Use this skill to make system-level design decisions and leave an inspectable path from drivers to evidence. It owns the architecture decision workflow, not the implementation of any subsystem.

Workflow

  1. Frame the decision. Establish the desired outcome, stakeholders, constraints, decision horizon, reversibility, affected systems, and evidence gaps. Do not invent scale, regulatory, latency, or ownership facts.
  2. Turn qualities into scenarios. Name the architecture characteristics that matter, express each as an observable scenario, prioritize them, and expose conflicts. Load references/architecture-characteristics-and-tradeoffs.md.
  3. Choose boundaries and shape. Compare styles, topology, and deployment granularity against the scenarios and team operating capacity. Load references/styles-topologies-and-granularity.md.
  4. Test ownership and coupling. Identify policy, data, change, runtime, and team boundaries. Treat a service split as a hypothesis, not a default. Load references/coupling-modularity-and-data-ownership.md.
  5. Make runtime behavior explicit. For asynchronous or distributed flows, specify authority, consistency, ordering, retries, duplicates, timeouts, partial failure, recovery, and reconciliation. Load references/distributed-workflows-and-consistency.md.
  6. Record and verify. Capture the decision and rejected alternatives in templates/architecture-design-brief.md and templates/tradeoff-record.md; use templates/architecture-review.md for challenge and sign-off. Define fitness evidence and drift response with references/evolution-fitness-functions-and-drift.md.
  7. Plan change without hiding execution ownership. Identify evolutionary slices, coexistence assumptions, and handoff conditions. Load references/migration-and-coexistence.md; route an approved transition to migration-engineering.
  8. Facilitate proportionately. Match review depth to blast radius, irreversibility, uncertainty, and cross-team impact. Load references/architecture-practice-and-facilitation.md.

Output Contract

Produce an architecture decision brief or review that includes drivers, stakeholders, constraints, prioritized scenarios, candidate options, explicit tradeoffs, boundaries and ownership, runtime and failure behavior, operational implications, decisions, evidence gaps, fitness checks, evolution slices, and named owners. State what is decided, what remains open, and which specialist owns follow-up work.

Ownership Boundaries

  • Reverse engineer an existing codebase or infer architecture from repository evidence with software-architecture-analysis.
  • Design interface contracts, schemas, compatibility, or API topology with api-design-and-evolution.
  • Design data platforms, data products, data models, or data governance with data-architect.
  • Implement services, integrations, transactions, or application code with backend-engineering and the relevant engineering owner.
  • Provision or operate cloud, network, CI/CD, containers, secrets, or observability substrate with platform-engineering and its tool owners.
  • Define security requirements, threat models, authorization, secrets, or security evidence with secure-software-engineering.
  • Author the durable ADR with adr-authoring; this skill supplies the architecture decision context and tradeoff analysis.
  • Execute an approved cross-system transition with migration-engineering; this skill decides whether the target shape and boundary are justified.
  • Model capacity, unit cost, load evidence, or SLO-cost tradeoffs with capacity-and-cost-engineering.
  • Design and exercise degradation, failover, restore, or recovery evidence with resilience-and-recovery.
  • Create structural diagrams with c4-diagramming or mermaid-diagrams.
  • Govern a technology portfolio, radar, or proportional technology governance path with technology-radar.

When Not To Use

Do not use this skill as a substitute for those specialist owners, as a code review workflow, or as a vendor/tool runbook. If the request is only one interface, data platform, implementation, infrastructure, security, capacity, diagram, ADR, or migration concern, load the narrower owner directly. If the architecture question depends on facts from an existing system, start with software-architecture-analysis and return here for a target decision.

Reference Guide

Load whenReference
Prioritizing qualities and comparing conflicting outcomesreferences/architecture-characteristics-and-tradeoffs.md
Comparing styles, deployment topology, or granularityreferences/styles-topologies-and-granularity.md
Testing modularity, coupling, boundaries, and data authorityreferences/coupling-modularity-and-data-ownership.md
Designing distributed workflows, consistency, or failure behaviorreferences/distributed-workflows-and-consistency.md
Choosing replication, partitioning, coordination, ordering, or transaction-isolation mechanismsreferences/distributed-workflows-and-consistency.md, then the DDIA mini reference; load the full reference only when deeper mechanism analysis is necessary
Defining fitness evidence, drift response, or evolutionary changereferences/evolution-fitness-functions-and-drift.md
Planning coexistence and handing execution to migration engineeringreferences/migration-and-coexistence.md
Running architecture workshops, reviews, and decision facilitationreferences/architecture-practice-and-facilitation.md
Checking provenance and the licensed-books transformation boundaryreferences/source-index.md

Keep the ownership split explicit when loading DDIA: this skill owns the system-level architecture decision and its tradeoffs; programming-principles supplies distributed-data principles and mechanism detail. Route message-handler implementation to backend-engineering, API or event contracts to api-design-and-evolution, data-platform design to data-architect, and exercised recovery evidence to resilience-and-recovery.

Completion

Stop when the architecture decision has an accountable owner, explicit alternatives and consequences, evidence or a named gap for each material claim, a verification path for prioritized characteristics, and specialist handoffs. Escalate rather than silently resolve missing authority, security, data ownership, or operational evidence.

Signals

GitHub stars
78
Forks
8
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
software-architecture
Source
github.com/magnus919/agent-skills