Show HN: Lelu – gate OpenAI agent actions on confidence and prompt injection
Category: security
Tags: authorization, ai-agents, security, prompt-injection, openai
Score: 7.5/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 7)
Lelu is an authorization engine for AI agents that gates OpenAI agent actions based on confidence levels and prompt injection detection. It provides a layered security pipeline including prompt injection filtering, confidence scoring, policy evaluation, and risk modeling, with support for human-in-the-loop review. The project is innovative in combining traditional authorization with agent-specific security concerns like prompt injection and confidence gating.
Target audience: backend devs, devops, security engineers
Repository: https://github.com/Lelu-ai/lelu · TypeScript · MIT · 36 stars
View on Hacker News