giving an AI agent permission to trade on your behalf also means accepting that it will make decisions when you're not there to watch every move that's why I've been paying attention to the risk controls behind @agenticscredit the risk engine has specific thresholds that determine what happens when a position starts moving against the agent. at 5% drawdown, the position size is cut in half. at 7.5%, positions are force-closed, and a 12% trailing floor stops the agent. I find this important because a trading strategy can perform well for several sessions and still run into a market it wasn't prepared for without clear limits, an agent could keep taking risks long after the original strategy has stopped making sense there's also a connection to the ACS, since trading activity feeds into the assessment of an agent's performance and risk management I'm interested in this approach because autonomy shouldn't mean allowing an agent to make unlimited decisions with capital. it should mean giving it room to operate within boundaries that protect against poor outcomes the ability to trade independently is useful, but knowing when an agent should stop is just as important.