The Anatomy of an Elite Prompt (CRISPE Framework)
Elite prompt engineers do not write unstructured paragraphs. They design prompts using modular components that establish persona, context, directives, and output formatting.
The CRISPE Framework
| Letter | Component | Purpose | Example |
|---|---|---|---|
| C | Capacity & Role | Who the AI should act as | "Act as a Principal Cloud Security Architect." |
| R | Recipient | Who the target audience is | "For junior DevOps engineers deploying to AWS." |
| I | Intent & Goal | What concrete outcome must be achieved | "Draft a hardened Terraform security checklist." |
| S | Scenario & Context | The background situation and environment | "We are migrating a legacy payment gateway to Kubernetes." |
| P | Parameters & Constraints | What to avoid, limits, and rules | "Max 10 bullet points. No theoretical fluff." |
| E | Examples & Format | The exact structure required | "Provide the output in a clean markdown table." |
Before & After Comparison
❌ Weak Prompt:
"Tell me how to make my website faster."
✅ CRISPE Prompt:
Capacity: Senior Next.js Performance Optimization Engineer.
Recipient: Frontend development team familiar with React 19.
Intent: Provide an actionable audit checklist to optimize Core Web Vitals (LCP, INP, CLS).
Scenario: An e-commerce store with heavy product photography experiencing 3.5s LCP on mobile 4G.
Parameters:
- Focus strictly on frontend changes (no CDN migration yet).
- Provide immediate quick wins (< 2 hours) and structural architectural fixes.
Format:
Use a 3-column table: [Metric] | [Underlying Cause] | [Actionable Fix with Code Snippet]