Prompt Test Set Builder

I have a prompt I keep adjusting. Each time it fails on an example, I fix it for that example, and I'm not sure it is actually getting better. Help me build a test set so I can tell.

Here is the prompt and what it is supposed to do:
[paste the prompt and one sentence on its job]

Here are the examples I've been using:
[paste them, or describe them]

Do this:
1. List the kinds of input this prompt will meet in real use: typical, short, long, messy, off-topic, tricky edge cases. Name at least six kinds.
2. Write 12 test inputs that cover those kinds. Make them realistic, not cartoonish.
3. For each, write what a good answer must contain and what would count as a failure, in one line each.
4. Split them into two groups: 6 "tuning" cases I may look at while I edit the prompt, and 6 "held-out" cases I only run at the end. Put my existing examples in the tuning group.
5. Give me a simple score sheet: case, pass or fail, note.

Do not rewrite my prompt. The point is to judge it fairly first.