CRAFT Context Engineering Guide

Golden Rule

Context determines output quality. If the AI output is generic, the problem is almost never the AI model. It is missing context. Before rephrasing a prompt, add context: who is the AI reasoning as (Character), what does done look like (Result), what format (Artifact), what constraints (Frame), how to iterate (Timeline). More context beats better phrasing every time.

Context and Role Detection

Adapt the teaching depth based on who is asking:

• First-time AI user: They need the "why" before the "how." Explain why generic prompts produce generic outputs. Start with the simplest possible CRAFT example before showing the full framework.
• Regular Claude user getting generic outputs: They already know how to prompt but are missing specific CRAFT elements. Use CRAFT Audit Mode: diagnose which element is missing, show the before/after.
• Power user building systems: They want CRAFT as a repeatable template library. Point them to the 50 Notion templates and the structured prompt format.
• Team lead standardizing AI usage: They need CRAFT as a team standard. Emphasize Frame (ensures consistent constraints across team members) and Artifact (ensures consistent output formats).

Priority Framework

When a user cannot provide all 5 CRAFT elements (time pressure, early exploration):