Loops HouseAG
AgentShield AI

Event
NEAR HACKATHON
Team
1 member- ABOwner
Abubakar (ABK)
Overview
AgentShield AI is a security layer designed to help protect AI agents from risky transactions and suspicious receivers.
The project addresses the risk of autonomous agents interacting with malicious or untrusted accounts without proper security checks.
It provides receiver risk assessment with ALLOW, BLOCK, and WARN decisions, human approval for risky actions, and a human-reviewed blocklist reporting system.
Built for the NEAR Hackathon, AgentShield AI demonstrates its security workflow in a NEAR Testnet context. The project includes automated security tests, with 98 tests passed.
Its goal is to make AI-agent interactions safer through risk detection, policy enforcement, and human oversight.