Prompt Engineering for ChatGPT - Vanderbilt University

Teaches systematic frameworks for constructing prompts through guided modules that decompose prompt engineering into discrete components (role definition, context injection, instruction clarity, output formatting). Uses a curriculum-based approach with worked examples and practice exercises to build mental models for how different prompt structures affect LLM behavior, enabling learners to move from trial-and-error to principled prompt design.

Teaches learners to recognize and classify recurring prompt patterns (e.g., few-shot prompting, chain-of-thought, role-playing, constraint-based prompting) through categorized examples and case studies. The curriculum maps these patterns to specific problem types, enabling learners to diagnose which techniques apply to their use case and understand the underlying mechanisms that make each pattern effective.

Teaches frameworks for assessing ChatGPT output quality across multiple dimensions (accuracy, relevance, tone, completeness, safety) and systematically using evaluation results to refine prompts. The curriculum provides rubrics and evaluation criteria for different task types, enabling learners to move from subjective 'this looks good' to structured assessment that identifies specific areas for prompt improvement.

Teaches learners to adapt general prompt engineering principles to specific domains (business, creative writing, technical documentation, customer service) through domain-focused case studies and examples. The curriculum demonstrates how to inject domain context, terminology, and constraints into prompts to improve relevance and accuracy for specialized applications.

Teaches techniques for maintaining coherent multi-turn conversations with ChatGPT, including context preservation, conversation state management, and progressive refinement through follow-up prompts. The curriculum covers how to structure conversation flows, handle context limitations, and use conversation history strategically to build on previous outputs.

Introduces learners to prompt injection risks, adversarial prompts, and techniques for hardening prompts against misuse. The curriculum covers how malicious inputs can manipulate ChatGPT behavior, common attack patterns, and defensive prompt design strategies to maintain intended behavior even when users attempt to override instructions.