← AI Agents & RAG learning pathAI Agents & RAG · Module 2 of 10
Prompt engineering
Write prompts that get reliable results.
📖 Key ideas
- Be specific: role, task, context, format, constraints.
- Give examples (few-shot) for consistent outputs.
- Ask for structured output like JSON.
- Break complex tasks into steps.
- Test prompts on many cases, not just one.
💡 Real-life example: A prompt is a job brief: a vague brief gets vague work, a clear one gets exactly what you need.
System promptInstructions setting behaviour
Few-shotIncluding examples in the prompt
Structured outputAnswers in a fixed format like JSON
Prompt templateReusable prompt with placeholders