Vague Prompts Waste Turns — Specify the Target

The most common prompting mistake: asking for something under-specified ('make this better', 'write about marketing') and getting a generic, average answer.

With no target, Claude optimises for the safest, most general response — which is exactly the one you didn't need. Specifying the audience, format, constraints, and success criteria collapses the space of possible answers onto the one you want.

Recognise this pattern whenever the output is technically fine but useless. The fix is almost always in the prompt, not the model.

Early brainstorming, where a deliberately open prompt is the point and you want breadth before you narrow.

Bad:  'Improve this copy.'
Good: 'Rewrite this landing headline for technical founders. Constraints: under 10 words, no buzzwords, lead with the outcome not the feature. Give 3 options and say which you'd ship and why.'

Asking 'explain caching' returns a textbook paragraph. Asking 'explain HTTP caching to a backend dev debugging a stale API response, focus on cache-control headers, 150 words' returns something you can actually use.

When you can't fully specify the target, ask Claude to specify it for you: 'Before answering, list the questions whose answers would most change your response, then answer them with reasonable defaults and proceed.' This turns a vague ask into a scoped one in a single turn.

• Blaming the model for a generic answer that a vague prompt guaranteed.
• Adding more words without adding more constraints — length is not specificity.
• Never stating the audience, which is the single highest-leverage detail you can add.