Prompt Evaluation Fixtures

Add a small fixture set for evaluating skill output quality.

The fixture set should include:

- prompts that should trigger this skill,
- near-miss prompts that should not trigger it,
- messy realistic prompts,
- expected output qualities,
- a simple scoring rubric.

This should help compare outputs before and after future changes to `SKILL.md`.

Avoid adding a heavy test framework unless it clearly pays for itself.