👋 Hello. This is Humanbound. We build open-source security testing for AI agents: prompt injection, tool misuse, data leakage, and the slow multi-turn manipulation single prompts miss. Apache 2.0, runs locally, no account needed. Here we'll share test findings, research we're reading, and what breaks when agents get tools. Pushback welcome. pip install humanbound https://github.com/humanbound/humanbound #introduction #infosec #AIsecurity #LLMsecurity #opensource