5 Templates

The chapter explains prompt templates as reusable frameworks for recurring tasks, built by combining structural elements, linguistic choices, and prompt patterns. Their main strength is the use of dynamic input parameters, which let one stable template adapt to many scenarios while keeping the core logic, constraints, and output shape consistent. Because templates are meant for repeated use, they should be stored externally, tested across varied inputs, and evaluated for quality, consistency, and failure behavior before deployment.

One major example shows how a template can turn feature requests into implementation plans in a repeatable way. By fixing the role, instructions, constraints, and output format while inserting changing feature details at runtime, teams can produce clearer and more comparable plans, reduce drift, and improve review quality. The chapter highlights how the output structure, concise wording, and role assignment work together to keep the response practical, scoped, and easy to use in engineering workflows.

A second, more advanced example uses a multistep workflow for risky legacy refactors: first, the model asks targeted questions to uncover hidden dependencies, migration risk, and rollback needs; then it uses those answers to generate a safer plan. This inversion-then-refinement approach shows why complex tasks benefit from multiple prompt patterns rather than a single pass. The chapter ends by emphasizing hands-on practice, encouraging readers to identify reusable prompt parts, separate fixed content from runtime inputs, and compare results for clarity, structure compliance, reusability, and effort.