dair-ai/prompt-engineering-guide

The Prompt Engineering Guide is a comprehensive educational resource that teaches people how to write better instructions for large language models, AI systems like ChatGPT, Claude, or Llama. The core problem is that these models are extremely sensitive to how you phrase a request: the same question worded differently can produce dramatically better or worse results. Prompt engineering is the skill of crafting those instructions deliberately to get reliable, high-quality outputs. The guide covers a wide spectrum of techniques with increasing sophistication. It starts with the basics, how to structure a prompt, what elements it should contain, and general tips for clarity. It then progresses to named techniques that researchers have developed: zero-shot prompting (asking the model to do something with no examples), few-shot prompting (giving it a small number of examples first), and Chain-of-Thought prompting (asking the model to reason step-by-step before answering, which dramatically improves accuracy on complex problems). More advanced topics include Tree of Thoughts (exploring multiple reasoning paths simultaneously), ReAct (combining reasoning with tool use), and Retrieval-Augmented Generation, known as RAG, where external documents are fed to the model alongside the question to improve factual accuracy. The guide also covers risks such as prompt injection, a type of attack where malicious text in the input hijacks the model's behavior, and covers practical applications in coding, data generation, and classification tasks. You would use this guide if you work with or build applications on top of large language models and want to get more reliable, precise outputs. Researchers, developers, and non-technical users who regularly use AI tools can all benefit. The content is written in MDX (Markdown with embedded components), which is a format used to build documentation websites.