AegisAgent was inspired by the growing risks in autonomous AI systems, where a single malicious prompt can manipulate agents into unsafe actions or data leaks. It acts as a real-time security layer that monitors inputs and agent decisions, detects prompt injection attempts and unsafe instructions, and blocks or modifies them before execution. We built it using a combination of rule-based filtering and LLM-powered analysis on a Python backend, creating a system that evaluates risk and ensures only safe, verified actions are carried out by AI agents.
Built With
- apis
- langchain
- python
Log in or sign up for Devpost to join the conversation.