How to Use Claude as a Software Architect: The Viktor Skill Guide

Writing the code is the easy part. Deciding how a system should be shaped before anyone writes it is where projects are quietly won or sunk. When people use Claude as a software architect with a quick prompt, they get a diagram-shaped answer with no reasoning about why it is built that way. A real AI software architect leads with trade-offs, because architecture is the discipline of choosing which problems you would rather have.

Viktor is that architect. He is a design persona you load once into Claude, ChatGPT, or any AI chat, and he refuses to hand over a single “best” design. He wants to know the scale you are building for, the team you have, and the constraints you are under — because the right architecture for a two-person startup is the wrong one for a bank, and vice versa.

Why generic AI architecture advice falls short

Unguided, a chatbot offers one confident answer dressed in fashionable technology, with no account of the trade-offs that make it a real choice. It ignores scale and constraints, recommending the same microservices-and-Kubernetes stack to a weekend project and an enterprise alike. A design with no stated downsides is not architecture; it is a wish list.

What changes with the Viktor skill

Viktor asks about scale, team size, budget, and constraints before proposing anything, then offers options with their trade-offs laid out — the cost, the complexity, the failure modes of each. He reasons explicitly about scalability, cost, and maintainability, and documents the decision and its rationale so the choice survives the people who made it. You get a defensible design, not a fashionable one.

What it actually produces

Architecture options with trade-offs, descriptions of system and data flow, technology selection with the reasoning attached, architecture decision records you can keep, and scaling and risk analysis. Describe what you are building and the constraints around it, and he maps the routes — and tells you what each one will cost you later.

How to get the most out of it

Give him the real constraints — expected scale, team size, budget, deadlines — because they are what separate good architecture from generic. Ask for options rather than a single answer, so you choose with the trade-offs in view. And ask him to record the decision and why, so future-you remembers the reasoning rather than re-litigating it.

Who this is for

Tech leads and senior engineers making design calls, founders deciding how to build before they commit, and architects who want a thinking partner. It works with Claude, ChatGPT, or any AI chat that accepts a system prompt. For the build itself, the tech & development skills collection covers coding, code review, and delivery — each a focused assistant rather than a general chatbot.