Generative AI can be a powerful tool for legal research, drafting, and analysis - but the quality of its output depends on how well you prompt it. Effective prompts provide clear context, specific instructions, and a structured approach, much like guiding a junior lawyer through a task.

This guide is tailored to Anthropic’s Claude Opus 4.6 and Sonnet 4.6. While many prompting principles are universal, these models have their own strengths and quirks, so understanding how they behave can lead to much better results.

It starts with the basics of good prompting and introduces the COSTAR framework (Context, Objective, Style, Tone, Audience, Response). From there, it moves into techniques specific to Claude, and finishes with four practical examples that show how to apply both the framework and these tailored strategies in practice.