This one walks you through system-level design decisions: monolith versus microservices, event-driven patterns, data consistency models, and how to actually decompose a system without creating a distributed mess. It includes decision trees for choosing architecture styles based on team size and consistency needs, plus templates for ADRs and architecture blueprints. The output guidelines are solid, they push Claude to recommend specific technologies rather than staying abstract, and to explicitly call out what not to build yet. Honestly refreshing to see a skill that treats premature distributed systems as the actual risk they are. The 2026 references cover platform engineering and AI-native patterns if you need current thinking, but the core content is the pragmatic stuff about Conway's Law and when a modular monolith beats microservices.
npx skills add https://github.com/vasilyu1983/ai-agents-public --skill software-architecture-design