The difference between a prompt and a good shared prompt: the second one works for other people too. For that, the changing parts – the text being summarized; the language being translated into – need to be marked as variables.
How variables work on PromptHub
You write variables in double curly braces. The wizard detects them automatically when you publish and lists them as input fields:
You are a precise translator.
Translate the following text into {{target_language}}.
Preserve tone and terminology. Output only the translation.
Text:
{{text}}
In each repo's playground the variables can then be filled in and tested directly – even by people who just discovered your prompt.
What makes a good variable
Speaking names: target_language instead of x – the name
doubles as documentation. As few as possible: every variable is a decision you
delegate to your users; anything that can be fixed belongs in the prompt itself.
Clear embedding: announce what a variable contains ("Text:" before the text block)
so the model never confuses instruction and input – which is also your best protection
against conflicting instructions inside the input text.
A typical beginner trap
Stuffing too much into one variable. If instructions exists as a variable,
your prompt is no longer a template but a blank page. A template's strength lies in the
fixed part – the variables are just the gaps.