Failure Mode 1: The Generic Identity. The identity could apply to any agent. "I am thorough, accurate, and helpful." Every agent should be thorough, accurate, and helpful. This identity differentiates nothing. The test: swap your identity onto a different agent in your system. If it still works, it's too generic.

Failure Mode 2: The Contradictory Identity. The components fight each other. A core belief that says "speed matters most" paired with a cognitive posture that says "examine every detail before proceeding." The agent will oscillate between rushing and overthinking, producing erratic behavior. Every component must point in the same direction.

Failure Mode 3: The Aspirational Identity. The identity describes what you wish the agent could do, not what it's designed to do. You write a metaphor referencing a visionary strategist, but the agent's actual function is formatting reports. The gap between identity and function creates confusing behavior. The identity should match the agent's real role, elevated but not fictional.

Failure Mode 4: The Instruction Identity. The identity is actually a set of instructions wearing a disguise. "My core belief is that I should always check for errors before responding." That's not a belief. It's a rule. Beliefs generate behavior across situations. Rules apply to specific situations. If your "core belief" could be expressed as an if-then statement, it's a rule, not a belief.

Failure Mode 5: The Orphaned Identity. The identity is beautifully designed but disconnected from the system. The agent has a strong cognitive posture, but no other agent in the pipeline accounts for it. A paranoid risk assessor that feeds into a synthesizer that ignores risk flags. The identity works internally but fails systemically. Every identity must be designed in context, aware of what it receives and what it produces for others.

You have just read five failure modes. The diagnostic below presents identity statements and asks you to identify which failure mode each one exhibits. This is pattern recognition practice. The faster you can spot a broken identity, the less time you will spend debugging behavior that traces back to a design flaw.

If you found yourself hesitating between Generic Identity and Instruction Identity, that is common. The two failure modes share a surface symptom (the identity feels flat) but differ in cause. A generic identity lacks specificity. An instruction identity has specificity but encodes it as rules rather than cognitive posture. The exercise that follows asks you to build an identity that avoids all five failure modes.

A Note on Iteration

Your first agent identity will not be your best. This is expected and fine. Identity design is a skill that improves with practice and operational feedback. You will write a v1.0 identity, test it against scenarios, discover gaps, and revise. The exercise in this module asks you to build a complete identity and then stress-test it. The stress test will reveal weaknesses. That's the point.

Do not aim for perfection. Aim for a complete v1.0 that's honest about its gaps and ready to evolve. Module 6 teaches evolution explicitly. But the instinct starts here: build, test, learn, revise. The version number is earned through operational experience, not editorial polish.