Foundational Techniques
The ten foundational prompt engineering techniques form the essential toolkit for working effectively with large language models. These techniques are widely applicable, relatively easy to implement, and produce measurable output quality improvements with minimal effort.
Who these are for: Developers, writers, analysts, and researchers who work with LLMs regularly and want to move beyond trial-and-error prompting.
Techniques in This Section
- Zero-Shot Prompting — Asking the model to complete a task with no examples
- Few-Shot Prompting — Providing examples to establish pattern and format
- Chain-of-Thought — Prompting the model to reason step by step
- Role-Play / Persona — Assigning an expert role to frame the response
- Structured Output — Requesting specific output formats (JSON, tables, lists)
- Negative Constraints — Specifying what the model should NOT do
- Context Injection — Supplying relevant background before the task
- Audience Specification — Defining who the response is written for
- Tone Control — Directing the register and emotional quality of writing
- Instruction Clarity — Writing unambiguous, well-scoped instructions
Each page follows the same seven-section format: definition → use cases → weak prompt → improved prompt → LLM outputs → verdict → pitfalls.