Prompt theory treats prompts as structured application components with distinct parts: task definition, instructions, context, examples, constraints, and output contracts. Key practices include defining narrow tasks with measurable success criteria, separating prompt sections clearly, placing stable rules in system messages and variable data in user messages, controlling context size, making constraints explicit with priority ordering, using few high-quality examples, designing explicit fallback behaviors, breaking complex workflows into prompt chains, versioning prompts like code, building evaluations before tuning wording, observing production behavior, and tuning model settings alongside prompt design. A practical checklist and a full support ticket routing example illustrate the framework end-to-end.
