super-brainstorm

To convert a vague product idea into a design spec, use an interview-driven planning process that forces deep context discovery and resolves key architectural decisions before coding. This validates scope and minimizes rework by producing an implementation-ready document.

The best way to plan architecture for a new feature is to perform a codebase-first context scan, then run a structured question flow to explore options and uncover constraints. This resolves unknowns and outputs a reviewed design specification.

Gathering requirements for a system refactor requires an ultrathink-based linear interviewing process that uncovers hidden assumptions and edge cases upfront. This enforces plan-mode entry to validate scope before implementation begins.

Yes, you can generate design specs for greenfield projects before writing code by applying an interview-first planning process. It explores architectural options, resolves key decisions, and saves a reviewed design spec to your docs/plans directory.

Codebase-first discovery improves requirements-gathering by performing a mandatory deep context scan of your existing code before asking questions. This reduces wasted back-and-forth by ensuring questions are highly relevant to your current architecture.

You should not use an interview-driven approach for simple component-level decisions that do not require upfront clarity or architectural planning. It is specifically built for complex refactors, new features, and system design discussions needing validated specs.