The 6 Core Elements of Effective Prompts

← Back to course

Most people treat prompts like search keywords. Professional prompt engineers treat them like structured specification documents. The difference in output quality is enormous.

Element 1: Role

The Role defines who the AI should embody. It has two layers: persona sets the domain and expertise level; skills refine specific capabilities.

Weak vs Strong Example

Weak: "Explain this code.
Strong: "You are a senior security-focused code reviewer. Audit this code and identify risks."

Element 2: Requirements

Requirements define what "good" looks like. It has three dimensions:

Content requirements: what topics to include and what to exclude
Format requirements: output structure (bullets, JSON, prose)
Quality requirements: reading level, tone, audience sophistication

Element 3: Task

The Task is the core instruction. Use precise verbs: analyze, critique, generate, rewrite, debug, summarize, classify, extract, plan, evaluate.

For complex goals, break tasks into numbered steps to force the model to think step-by-step.

Element 4: Examples

Examples are the most underused element, and often the highest-leverage addition. They work in three ways:

Successful examples: show the expected format and style
Failure examples: explicitly name error patterns to avoid
Format templates: provide a precise output structure

Few-Shot Prompting

This technique is called few-shot prompting. Research shows it produces more consistent outputs than bare instructions, especially for tasks with strict formatting requirements.

Element 5: Constraints

Constraints define boundaries the model must respect. They serve three purposes:

Hard rules (red lines): non-negotiable constraints like word count or no medical advice
Preferences: soft guidance for when the model has agency, like prefer active voice
Risk avoidance: safeguards against problematic outputs like don't speculate beyond what's provided

Element 6: Process

The Process defines the sequence of steps the model should follow before producing the final output. It's the most powerful element for handling complex tasks.

This technique is called Chain-of-Thought prompting. Forcing the model to externalize its reasoning step-by-step dramatically reduces errors and makes output verifiable.

With vs Without Process

Without process: "Is this business idea viable?" — produces opinion without logic.

With process Step 1: Identify the target customer and their core problem
Step 2: Assess whether the problem is mainstream or niche
Step 3: Evaluate whether the solution is better than existing alternatives
Step 4: Identify the top three risks
Step 5: Provide a judgment with reasoning based on the above analysis

Full Example: Using All 6 Elements

Role: "You are a content strategist writing for busy professionals.
Requirements: 400 words, H2 section headings, direct tone, each section needs one real case study.
Task: Explain why most people fail to build habits and the three most effective strategies.
Examples: [paste a sample paragraph showing the expected style]
Constraints: Do not reference Atomic Habits; no motivational language; flag unverified claims.
Process: 1) One sentence stating the problem 2) One paragraph explaining the difficulty 3) Three H2 strategies 4) One-sentence summary.

When to Use Each Element

Simple tasks (translation, typos): Task alone is enough
Medium tasks (descriptions, concept explanation): Role + Task + Requirements
Complex tasks (strategic analysis, long docs): All six elements
Production systems (real features in a product): All six elements, with special attention to Constraints and Process

Key Takeaway: Start by adding one element you usually skip. Most people skip Examples and Process — just those two will dramatically improve results. Structure your prompts and you'll consistently get outputs that need far less correction.