AI Agents / Web3

Guardrail

"An AI trading agent that can't go rogue."

Role

Full-Stack Engineer

Client / Project

Emmanuel Ocheme

Technology Stack

TypeScriptviemSolidity

Links

Code

Guardrail Case Study

The Challenge

An autonomous on-chain trading agent is hard to trust: you want plain-English control but also hard limits the agent can't bypass, and a way to stop it without race conditions.

The Solution & Architecture

A vault contract on HSK Chain enforces the owner's risk limits on-chain: the agent can trade but never withdraw, every trade is checked against the owner's limits, its reasons are logged on-chain, and either the agent or the owner can hit the kill switch.

Impact & Results

A plain-English-steered on-chain trading agent with enforced risk limits and a kill switch, built for HSK Chain Horizon.

Vault contract enforces owner's risk limits on-chain
Agent can trade but never withdraw
On-chain kill switch held by either side