Prompt Repetition

This is a bounded decision rule for when duplicating your prompt 2x or 3x actually fixes accuracy problems on lightweight models like Haiku or Flash, and when you should restructure instead. The fit is narrow: long-context retrieval where the question gets buried, options-first multiple choice, or position-sensitive lookups in inventories. It walks you through classifying the prompt shape, checking if you can afford the input token overhead, and picking the smallest intervention before you burn context budget on full duplication. The skill explicitly routes you away from repetition when the real problem is retrieval quality, reasoning depth, or tool-heavy orchestration. Useful when someone says the model forgot the instruction at the end and you need to know if repeating is a cheap fix or just multiplying the wrong prompt.

Install to Claude Code

npx -y skills add akillness/oh-my-skills --skill prompt-repetition --agent claude-code