Edge Case Persona Behavior

Agents reveal their true personality at the edges — when input is ambiguous, when they are asked to operate outside their domain, when two of their values conflict. A sales-focused agent asked to deliver bad news. An analytical agent asked to make a decision without sufficient data. A consensus-seeking agent asked to override another agent's recommendation. These edge cases are where persona specifications prove their depth or expose their shallowness.

Edge case behavior must be explicitly designed, not left to inference. For each persona, define behavior for five standard edge cases: insufficient data (what does the agent do when it cannot produce a confident output?), domain boundary (what does the agent do when the request falls outside its expertise?), value conflict (what does the agent do when two of its core values produce contradictory guidance?), error discovery (what does the agent do when it realizes its previous output was wrong?), and human override (what does the agent do when a human disagrees with its recommendation?).