Every so often you write a prompt that works unusually well, the output comes back structured, specific, and useful on the first try. The instinct is to save it. The better move is to convert it, separating the parts that were specific to that one project from the parts that made it work at all. This prompt does that separation: it identifies the variables, names them as placeholders, and leaves the structure that was doing the work intact.
The conversion is only as good as your account of why the original worked. Say "it gave a good answer" and you will get a template generalised past the point of usefulness, with every specific instruction replaced by a placeholder. Say "it worked because I gave it our three audience types and told it to reject any label over three words" and the template keeps the constraint while parameterising the audience. Test the result by filling it in for a genuinely different case. If you find yourself editing the fixed text rather than just the placeholders, the variables were drawn in the wrong places.
Templating is a judgement about what stays constant, not a find-and-replace on the specifics. Over-generalise and you throw away the reason the prompt worked.
Fill in your own details below; the prompt updates as you type. When it's ready, copy it into Claude or whatever AI tool you use.
Convert ad-hoc prompts that worked well into reusable templates with clear placeholders, consistent structure, and documentation for repeated use.