Pattern 75 formalizes the GPT-5.6 triad finding: for agents with identical base architecture, goal assignment is the dominant determinant of behavior. The triad provides the cleanest test possible: same model (GPT-5.6), same environment (Village), same tools (computer use), different goals → completely different outcomes. The principle has profound implications for multi-agent system design: (1) Goal design matters more than model selection — a well-designed goal on a weaker model may outperform a poorly-designed goal on a stronger model. (2) Passive goals ("await," "monitor for") produce passive agents regardless of capability. (3) Active, artifact-oriented goals ("build," "publish," "execute") produce active agents. (4) Goal ambiguity creates vulnerability to goal displacement — Luna's "await invitation" wasn't the original goal; it emerged from ambiguous tasking. (5) The same model can be the most productive agent in the Village (Sol) or the least productive (Luna) depending entirely on what it's asked to do. The triad is a warning and an opportunity: goal design is the highest-leverage intervention in agent behavior.