Adversarial Prompt Testing Framework

# Adversarial Prompt Testing Framework

You are a QA engineer specializing in AI systems. Build an adversarial testing framework for the prompt below, designed for [APPLICATION TYPE].

**Prompt under test:** [PASTE YOUR PROMPT HERE]

## Adversarial Test Categories

Generate test cases for each category:

**Boundary inputs:** Empty input, single character, maximum length input, input with only whitespace or special characters

**Ambiguous inputs:** Input that could be interpreted multiple valid ways, input that matches multiple categories, underspecified requests

**Contradictory inputs:** Input that conflicts with the system prompt constraints, input that asks for something the persona would refuse, input that mixes two incompatible formats

**Injection attempts:** Instructions embedded in user input, role override attempts, fake delimiter injection

**Edge case data:** Unusual languages, mixed scripts, numbers formatted differently, dates in multiple formats

## Test Case Documentation

For each adversarial input, document:
- Category and rationale
- Expected output behavior
- Actual output (fill in after testing)
- Pass / Fail verdict and severity if failed

## Failure Remediation

For each failure found, specify the minimum prompt change required to fix it without breaking other test cases.