The problem DarkAgent solves

As AI agents become more active in onchain finance, users need a way to let agents act without giving them unrestricted wallet power. Today, an AI trading bot, social signal, or shared Blink can trigger risky or oversized transactions, route through untrusted protocols, or buy unsafe assets. DarkAgent solves this by acting as a policy firewall between agent intent and wallet execution. It ensures every AI-generated action is checked against user-defined limits before funds move.

Challenges I ran into

• Designing a product that feels like a real agent execution console instead of a static hackathon mockup.
• Translating abstract policy controls into a simple, understandable UX for judges and users.
• Making the wallet interaction feel real while keeping the execution flow safe for demo purposes.
• Coordinating policy enforcement, Blink parsing, wallet approval, and proof generation into one believable end-to-end flow.
• Balancing security language, DeFi functionality, and AI-agent use cases without overwhelming the user.