Context strategy is the design layer above individual prompt decisions — the overall plan for how information is assembled and delivered to the model at each point in a workflow. A context strategy might specify: system prompt comes first, followed by retrieved documents ordered by relevance, then conversation history trimmed to the last five turns, with user data appended as structured JSON. It considers what the model needs to know, what it can safely ignore, and how to arrange information so the model processes it reliably. For behavior architects, context strategy is where product goals meet technical constraints: it requires understanding both what the model is good at reasoning over and what information actually makes a difference to the quality of its responses.