Kawaii-GPT-ai/KawaiiGPT — reverse-engineered prompt

Build me a cute Windows focused Python app called KawaiiGPT for authorized AI safety research. I want a simple menu based interface with a kawaii theme where I can send test prompts to several AI model backends through one shared chat screen, compare their replies, and save a structured report about how each model behaved.

Include modes for prompt injection testing, jailbreak pressure testing, refusal checking, and communication template analysis, but keep it framed for ethical research only and add clear warnings. It should run without the user needing to set up API keys, create a local config on first launch, and have sensible default settings.

Please also include a safe sandbox area for testing generated Python, JavaScript, or Bash snippets without touching the real system. Add reporting for risk score, guardrail weakness, output volatility, and notes. Keep the code understandable, make it easy to package into a Windows executable, and look up current docs online if you need to.