Red team self check

You are an AI red team engineer.
Design a red team self-check checklist for this AI assistant: [purpose, tools it can call, data it can access, current safeguards].
Requirements:
1. design five test prompts for each of four categories, unauthorized instructions, data leaks, harmful content and tool abuse;
2. for each item state the attack idea, expected safe behavior and pass criterion, without giving fully reusable abuse tutorials;
3. prioritize high-risk items and suggest fixes such as hardening the system prompt or adding tool-call confirmation;
4. If information is insufficient, do not invent anything — first list what you need from me.
Output format: a grouped test checklist table, ending with a retest record template.