Use when the user has a plan, idea, strategy, pitch, launch, architecture, or decision they're about to commit to — any moment where nodding along would let them walk into a failure they could have spotted. Steel-mans the opposition, exposes hidden…
Use when the user faces a decision with multiple options — "should I pick X or Y", "which approach is better", "help me choose", comparing tools/stacks/strategies/vendors, or any moment where answering with the first plausible option would skip better…
Use when the user invokes "best practice", "standard approach", "everyone does X", "the industry does Y", or when the default answer smells like received wisdom rather than evidence. Also use when hunting for contrarian advantage — "why does nobody do X", "is…
Use when the response will contain factual claims the user might act on — numbers, dates, API shapes, library behavior, historical facts, "how does X work" answers, or anything where a confident wrong answer would waste the user's time or cause them to trust…
Use when a problem mixes strategy and tactics — "what should I do about X", "how should I approach Y", career/business/product questions, situations where the answer at one level (industry trends) has to connect to the answer at another (what to do this…
Use when the user has a concrete GOAL they want help pursuing — "I want to achieve X", "help me hit Y", "how do I get from here to there", launching something, driving a metric, reaching a milestone. Clarifies the objective, models the current situation,…
Use before writing any long or multi-part response — documents, essays, strategy write-ups, architecture explanations, or any answer where jumping straight to prose would produce rambling, incoherent, or out-of-order output. Decomposes sub-goals, maps…
Use when the user wants to LEARN a topic, not just get an answer — "explain X to me", "teach me Y", "I want to understand Z", "help me wrap my head around", onboarding to an unfamiliar domain, or any moment where a one-shot explanation would leave gaps.…