how

Why we included this project

Getting a new engineer up to speed on an unfamiliar codebase usually means long walkthroughs and a lot of patience. This Cursor skill automates part of that: the `/how` command turns a plain-language question like "how does message virtualization work?" into a structured architectural explanation with sections for overview, key concepts, how it works, where things live, and gotchas. Questions that span multiple files or services get split into parallel exploration angles handled by subagents, whose findings a synthesis agent merges into one coherent answer; simple questions run end-to-end in a single agent. A critique mode goes further, spawning independent critics across different models to surface architectural problems after the explanation is done. Teams working in Cursor who want onboarding-level code comprehension without hand-building their own prompt chains will find the routing logic and prompt templates worth studying or adapting.