AISecurityLab/hackagent — reverse-engineered prompt

Build me a Python toolkit called HackAgent that helps people safely red team their own AI agents and find issues like prompt injection, jailbreaking, goal hijacking, and tool misuse. I want it to work as both a CLI and a Python SDK, with a nice terminal experience that shows attacks running in real time and gives clear results at the end.

It should work locally without needing an API key, and let me test agents built with common setups like OpenAI, LangChain, LiteLLM, and Google ADK. Include a modular attack pipeline with a generator, a judge, the target agent, and support for built in benchmark datasets plus custom file or HuggingFace datasets. Save local results in SQLite and give me a simple web dashboard to review runs, and if an API key is present, allow syncing to a hosted app too.

Please include basic tests, docs, and responsible use messaging so it is clearly for authorized security testing only. Look up current docs online if you need to.