The CRAFT and CHECK Frameworks
About this resource
Everyday AI use needs two habits, not one. Better instructions going in, and honest verification coming out. Most teams learn the first and skip the second, which is how a fluent, confident, completely wrong answer ends up in front of a client. This free one-page handout puts CRAFT and CHECK side by side and explains why they only work as a pair.
CRAFT covers what goes into a prompt so the output is worth having. CHECK covers what has to be examined before that output counts as work. Both are widely used community mnemonics rather than anything we invented, and we teach them because they are short enough to remember under time pressure, which is the only property that matters in a mnemonic. The page includes what each letter stands for and the specific question it prompts, in language a new hire can use on their first day.
One for the input, one for the output. Teams that learn only the prompting half produce better-written mistakes, which are considerably harder to catch than obvious ones.
How to use it
Use it as the anchor handout when a team first gets access. The prompting half produces the immediate satisfaction and the verification half produces the safety, and teaching them together prevents the most common sequencing mistake in AI training. If you can only get one habit to stick, make it CHECK. A poorly prompted answer wastes a minute; an unverified one can cost considerably more.
Inside the download
A prompting discipline and a verification discipline, presented as one habit rather than two.
Five letters each, short enough to use without looking anything up.
Why a well-prompted wrong answer is more dangerous than a badly prompted one.
A single page a new hire can be handed on day one.
Common questions
CRAFT is a widely used community mnemonic for structuring an AI prompt so the output is worth having. It prompts the user to supply the elements a model cannot infer, including role, context, and desired format. It is not proprietary to any single firm and is taught broadly across the field.
CHECK is a community verification mnemonic covering the five things worth examining before AI output is used: the claims it makes, whether anything could be fabricated, whether the reasoning holds, whether the model had the context it needed, and whether the topic falls into a knowledge gap such as recent, proprietary, or industry-specific information.
Yes, and skipping verification is the more common and more expensive error. Better prompting improves the quality and the confidence of the output simultaneously, which means a well-prompted wrong answer is harder to spot than a badly prompted one. The two habits are taught together for that reason.
Free. No email list, no follow-up sequence.
More from the resource library, chosen because they pair well with this one.
A five-step human-in-the-loop verification discipline you can run on any AI response in under sixty seconds.
Six before-and-after prompt pairs that show exactly what changes when a vague request becomes a specific one.
Where human judgment still matters, three levels of stakes, and what review actually has to mean at each one.
Tell us a little about yourself and this document is yours. We don't have an email list. We collect this so we know who's finding our resources useful, and that's it.
You're downloading
this document
Your information is kept confidential and never shared with anyone.
