◇ Could this help me?
Yes, and this is one of the highest-value items in the library relative to its size. Vague instructions plus a confident agent is the single most common way people end up with a large amount of code that solves a slightly different problem, and forcing the ambiguity to surface as questions before implementation is far cheaper than discovering it afterwards. The third sentence is what makes it practical rather than exhausting, because an agent that can read the codebase should not be spending your attention on questions the code already answers. General practice, and the only calibration needed is that on a genuinely small task the interview is overhead, so reach for it when the plan is big enough that being wrong would cost real work.
Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one by one. If a question can be answered by exploring the codebase, explore the codebase instead. Do not write any code until the interview is finished and you have summarized what we agreed, including the decisions we made and the alternatives we rejected.