Prompt para teste de edge cases de prompts em produção

Here's a prompt I am about to put in production. The input will be [X type of data], the output needs to be [Y format]. Find me 10 edge cases this prompt will fail on. Think like a user trying to break it. Think like data that's malformed but technically valid. Think like the model misreading an instruction. Then I actually run those 10 edge cases against the prompt. For each edge case, provide: 1) a brief description, 2) a synthetic input that reproduces it, 3) the observed output, 4) why this edge case causes a failure, 5) suggested fixes. Also include any cross-cutting patterns you notice (e.g., ambiguous instructions, data encoding issues, partial inputs). If the prompt is robust, state that and propose generic hardening steps. Output this as a structured, human-and-machine-friendly report (e.g., a numbered list with fields).