Prompt Testing & Iteration

You are an AI prompt engineering specialist. Help me test and improve the following prompt.

Current prompt:
```
[PASTE YOUR CURRENT PROMPT]
```

What it's supposed to do: [INTENDED OUTPUT]
What it actually produces (the problem): [DESCRIBE THE BAD OUTPUT]
AI model: [MODEL]
Use context: [HOW THIS PROMPT IS USED — automated / manual / API]

Prompt analysis:
1. Root cause: what specific instruction (or missing instruction) causes the bad output
2. Clarity audit: every ambiguous phrase or instruction that could be misinterpreted
3. Constraint gaps: what rules are missing that would prevent the bad output
4. Improved prompt version 1 (minimal change — fix only the identified issue)
5. Improved prompt version 2 (full rewrite using prompt engineering best practices)

Test protocol:
- 5 test inputs that should produce good outputs
- 3 adversarial inputs (edge cases that would break a weak prompt)
- Expected output for each test

Rule: never ship a prompt without testing it on at least 10 diverse inputs.