🦊

PROMPT ARENA

What is Prompt Engineering?

Prompt engineering is the science of programming Large Language Models (LLMs) via natural language. Because LLMs process text sequentially using self-attention mechanisms, prompt structure directly changes the statistical weights of the generated tokens. By controlling these attention pathways, we enforce security boundaries, format strict outputs, and guide logical reasoning.

What We Are Going To Do

You will step into the role of an AI Systems Engineer and solve 6 real-world challenges. For each challenge, you will design the prompt template. When you click run, your prompt will be tested by an automated validation agent against strict parsing checks, prompt injection attacks, and token-efficiency standards in real-time.

The 6-Level Curriculum

L1
System vs User Prompts
Isolate permanent application rules (System Prompt) from raw user queries to prevent context hijacking.
L2
Context Isolation & XML Tags
Draw clear structural boundaries around untrusted inputs using tags to block injection attacks.
L3
Prefilling Assistant Responses
Guide the first token of the output to bypass chatty greetings and guarantee raw JSON.
L4
Chain of Thought Reasoning
Give the model a scratchpad to reason step-by-step, reducing logic and reasoning errors.
L5
Few-Shot Prompting
Provide high-quality input-output examples in the context to teach complex classification patterns.
L6
Citations & Refusal Boundaries
Force quote extraction evidence loops to stop factual hallucinations and QA fabrication.
Developed by Orbit AI for Deep Learning Indaba x Zimbabwe 2026