AI coding agents show up in DevOps all the time, but they always hit a wall. They lose context, have no feedback loop for pipeline failures, and forget everything about your app between sessions. When we tried to use an agentic tool to perform a complex infrastructure migration, we hit all three problems immediately. Instead of using a bigger model with a larger context window or a multi-agent framework, we built a lightweight pattern around Agent Skills to more effectively guide it on completing the task. This talk walks through what we tried, what broke, and the simple pattern that made it work. AI tools are very powerful when used effectively, and if you’re experimenting with generative AI, you’ll walk away with concrete principles and a reusable approach that you can try on Monday.