Loops HouseAgent Shield

NEAR HACKATHON
Links
Team
1 member- KUOwner
kufreAbasi
Overview
AI agents are starting to hold wallets, move money and act on their own, but there is no simple way to tell whether an agent is safe to trust. Agent Shield is a security and trust platform for AI agents. Developers connect their agents with a lightweight SDK, and Agent Shield records what each agent does, flags risky or suspicious behavior, and shows the results in a clear dashboard written in plain language, so anyone can understand it, not just security experts. It also includes an open threat repository: anyone can submit a threat report, and anyone can browse the public repository, so the whole community learns from each attack instead of every team finding out alone. The platform is built with Next.js, SQLite and Argon2id, and was tested with three mock agents that show normal and malicious behavior. It fits NEAR's AI Security & Defense track because agents on NEAR are being given autonomous control over money, and that only works if people can monitor them and trust them.