Help me understand: depthfirst seems to be bigging up their “security agent” here, but is it not just prompt engineering + writing skill files? What goes into producing a “security agent” beyond this? Feels like they’re really gussying up a process that is ultimately just LLM usage