A strong prompt has four parts: Context, Align, Goals, Examples. Assemble one to decode a tender, load a ready recipe, then copy it into your sanctioned AI tool. The skill is asking well — and making the AI check in before it runs.
C · Context — everything it needs: who it is, what it's reading, the rulesA · Align — make it play back its plan and wait for your goG · Goals — what "done" is, plus what it must NOT do and must FLAGE · Examples — a sample to copy
Load a recipe
C · ContextAll of it — who the AI is, what it's reading, the situation and the rules.
A · AlignMake it play back the plan and wait — correct the approach before the output.
G · GoalsThe single, specific job — and its guardrails live right here.
E · ExamplesShow one sample line so it copies your format exactly.
Your CAGE promptWeak
A weak prompt (don't do this)
"summarise this tender" — no context, no goal, no example, no guardrail, and no chance to correct it before it runs. You'll get a vague wall of text, it may invent requirements, and you can't defend what you submit. CAGE is what turns AI from a gamble into a tool you can stand behind.