A library tutor-bot prompt tells the model to refuse homework cheating and instead offer study strategies. What prompting basic is applied?

A library tutor-bot prompt that refuses homework cheating and instead offers study strategies applies boundary instructions. Civic educational tools must respect academic integrity while still helping residents learn. Encoding refusal of graded-work completion and redirecting toward strategies, explanations of concepts, and resource pointers is a prompting basic that defines acceptable assistance.

Persona instructions in the prompt that refuse cheating help by adopting a supportive tutor voice fail because they mislabel the prompting basic: a supportive voice sets persona, while refusing graded-work completion and steering toward study support is a boundary instruction that defines acceptable assistance. Formatting instructions in the prompt that refuse cheating help through structured steps and summaries fail because they mislabel the prompting basic: organizing output into steps and summaries is formatting, while refusing to complete homework is a boundary instruction that sets limits on assistance. Example dialogues in the prompt that refuse cheating help while modeling practice problem explanations fail because they mislabel the prompting basic: showing model explanations is use of examples, while the rule to refuse cheating and redirect to study strategies is a boundary instruction.

Exam caveat: boundaries should be specific ("do not complete graded assignments") and paired with positive alternatives ("explain the method; ask guiding questions"), not merely vague "be ethical" slogans. Operational check: test the bot with cheating-request prompts, verify refusal plus constructive redirect, log escalations to human tutors, and review prompts with education partners before public launch.