CoStar vs Other Prompting Frameworks: When to Use Which

The prompt engineering landscape has produced many structured frameworks, each optimized for different scenarios. CoStar (Context, Objective, Style, Tone, Audience, Response) emphasizes communication-quality outputs — ideal for content generation, marketing copy, and any output where style and audience matter. CRISPE (Capacity, Insight, Statement, Personality, Experiment) emphasizes role definition and iterative exploration — useful for creative and open-ended tasks. RISEN (Role, Input, Steps, Expectation, Narrowing) emphasizes process — ideal for tasks with clear sequential steps like data transformations or analysis pipelines. RTF (Role, Task, Format) is a minimalist three-slot framework — good when you don't need heavy context and just want clear role plus clear task plus clear format. TAG (Task, Action, Goal) is similar but emphasizes the why of the task — useful for persuasion or strategic communication. For code generation specifically, frameworks like Anthropic's XML-tagged prompt structure with <task>, <context>, <examples>, <constraints> sections often outperform generic CoStar because code tasks benefit from explicit input/output examples more than from tone specification. The selection criterion: match the framework to the dominant failure mode of your current task. If outputs sound generic or miss the audience — CoStar. If outputs skip process steps — RISEN. If outputs are off-tone — CoStar. If outputs ignore constraints — frameworks with explicit constraint slots. The deeper skill isn't memorizing frameworks — it's diagnosing what your prompts are missing and selecting the template that supplies that element.