Every vague word is a decision you handed to the model. Take them back.
The cost of vagueness
Words like some, various, appropriate, properly and etc. read as instructions but carry no information. The model still has to act, so it guesses, and its guess is an average of everything it has seen, which is exactly the generic output you were trying to avoid.
The swap
Trade each hedge for a number, a name, or a rule. 'Keep it short' becomes 'under 120 words'. 'Some examples' becomes 'exactly three examples'. 'Handle errors properly' becomes 'return null on a parse failure and log the offending line'. Deleting a hedge without putting a fact in its place just leaves a shorter vague prompt.
When vagueness is right
Leave a term open only when you genuinely want the model's judgement, and say so: 'pick a tone you think fits a technical audience'. Deliberate latitude reads very differently from an unfinished sentence.
Try it
Replace every vague term with a concrete fact. Aim for a score of 85+.
74structure
Target: 85. Keep working the prompt. The score moves as you type.
⌃Constraints: State limits and do/don't rules, e.g. "Keep it under 200 words. Do not invent APIs."
⌃Output format: Specify the exact shape, e.g. "Return JSON: { title, steps[] }" or "Use a markdown table.".
⌃Examples (few-shot): Add one input→output example to anchor tone, format, and edge handling.