Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Negative Prompting
Negative Prompting
Steer output by naming what to avoid, each paired with the positive alternative to do instead.
ROLE: You are an editor who improves results by defining clear no-go zones, knowing what you leave out matters as much as what you include.
CONTEXT: For [TASK], previous outputs kept showing unwanted traits such as [BAD_PATTERNS], and I want those eliminated without losing required content.
TASK:
1. Turn each unwanted trait into an explicit do-not rule.
2. For every do-not rule, state the positive alternative to do instead, so nothing is left to guesswork.
3. Generate the output while honoring both the exclusions and the alternatives.
4. Scan your own draft and remove any instance of a forbidden trait.
5. Confirm each exclusion was respected.
CONSTRAINTS: Avoidance rules must be specific, not vague. No cliches, filler, or hedging. Never drop required content just to satisfy a do-not rule; if avoiding something makes the task impossible, explain the tradeoff. Note any place you had to resist a common default.
OUTPUT FORMAT: Section DO-NOT / DO-INSTEAD as paired bullets; section OUTPUT; section AVOIDANCE CHECK listing each forbidden trait and confirming it is absent.
Produce it, avoiding the above, for: [TASK]