Skip to main content
PROMPT-CACHING5 MIN READ

Walk a support prompt into a cacheable shape

Refactor a support-triage prompt by choosing the next cache boundary decision at each step.

The slow triage call A 5,800-token support prompt repeats policy, queues, rubric, and tools on every ticket, but starts with the changing ticket facts. The stable work is present; it is just in the wrong order. Stable-prefix walk Frame -> Tail -> Measure Preserve the reusable frame, append live facts clearly, then verify cache hits instead of guessing. Mixed prompt ticket facts, policy, tools, and rubric interleaved The model sees the same long prefix across tickets and still gets fresh facts before answering. Do not trade answer clarity for cacheability; get both with labels and order. 01 Frame 02 Tail…

Read the full lesson

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in

← Back to library
Contact us