tkostanyan/llm-testing-kit ? reverse-engineered prompt

Build me a Python app that can red team tool using AI agents for prompt injection.

I want to point it at different kinds of agents, like an email helper or a document search assistant, and have it run a set of attack prompts against them, then compare how well a few layered defenses work. It should include a couple of example targets so I can test it right away, plus a way to plug in my own target later.

Please also add a small set of normal benign tasks so the app can measure how often the defenses get in the way. At the end, I want it to generate a clear report with charts and tables that show attack success rate, false positive rate, and latency, and save both the data and a PDF report.

If needed, look up current docs online for the model provider or PDF tooling.