📃 System Prompts

Base Instructions

Foundational rules guiding AI’s overall behavior

Define tone, style, safety boundaries

“Be concise, factual, and polite.”

Behavior Constraints

Hard limits or guardrails on AI outputs

Prevent harmful, biased, or off-topic responses

“Do not generate personal information”

Context Injection

Embedding relevant session or environment info

Ensures AI responses consider current context

User role, recent actions, system state

Few-shot Examples

Sample interactions to illustrate expected output

Helps AI mimic desired patterns

Provide 3 Q&A pairs demonstrating style

Output Formatting

Specify structured output formats or templates

Enforces consistent data or UI integration

“Return JSON: {‘summary’: text, ‘score’: number}”

Fallback Instructions

What AI should do if uncertain or off-topic

Improves reliability and safety

“If unsure, respond with ‘I don’t know’.”