A prompt framework is a reusable recipe for composing prompts instead of designing each one from scratch. It organizes principles such as clarity, relevant context, iteration, and conversational continuity so prompts can be applied more consistently across tasks, audiences, and domains. section 7.5 section 7.4
Common frameworks emphasize different kinds of control:
- ReAct (reasoning and acting) alternates model reasoning with external actions such as tool or application programming interface (API) calls, then uses the returned results. section 7.5
- MECE (mutually exclusive, collectively exhaustive) organizes an answer into non-overlapping categories that together cover the problem space. section 7.5
- PACT structures a prompt around persona, action, context, and tone, while WISER uses who, instruction, subtask, examples, and review. section 7.5
Table 7.4 Example prompting frameworks for systematic prompting
| Framework |
Description |
Example |
| ReAct (reasoning + acting) |
Combine reasoning steps with tool or API calls |
“I need flight options for Austin → call flight API → evaluate results → recommend itinerary.” |
| MECE |
Organize outputs into non-overlapping, complete categories |
“List customer segments using a MECE structure: Demographic, Behavioral, Geographic, Psychographic.” |
| PACT |
Template: persona, action, context, tone |
“You are a luxury fashion marketer. Draft Instagram captions for sustainable leggings in an aspirational tone.” |
| WISER |
Template: who, instruction, subtask, examples, review |
“You are a product manager → Summarize product launch → List key features and reactions → Provide sample summaries → Review for completeness and clarity.” |
Compare each framework’s structure, purpose, and example prompt in one view. section 7.5
Frameworks can improve alignment between user intent and model output, but they are not universal requirements: capable models may need less prompting structure. Effective use still requires evaluating sample responses, scoring results, and iterating; some tasks may be better served by post-training or a hybrid approach. section 7.13 section 7.4
