Name the schema before building the agent — a reliably…
Before building any agent, explicitly name the schema it operates on by writing what the agent assumes about how the world works, because unexamined schemas produce systematically wrong outputs despite reliable execution.
When This Fires
- Before designing any new behavioral agent — as a mandatory pre-design step
- When an agent fires reliably but produces bad outcomes — the schema may be wrong
- When importing someone else's behavioral system (a book's advice, a coach's recommendation) — their schema may not match your context
- During periodic agent reviews when reassessing whether the underlying model still holds
Common Failure Mode
Importing agents without examining their schemas: "Tim Ferriss batches email twice a day, so I should too." Ferriss's schema: his business doesn't require real-time email responsiveness, and batching frees cognitive resources for creative work. Your schema may be different: your role may require fast responses, and the anxiety of unanswered email may consume more resources than processing them does. The agent is the same; the schemas are different; the outcomes will be different.
The Protocol
(1) Before building an agent, write one paragraph: "This agent assumes that..." — state the model of the world that makes this agent's design rational. (2) Evaluate: is this model accurate for your specific situation? Not "is this generally true?" but "is this true for me, now?" (3) If the schema is clearly correct → build the agent. (4) If the schema is questionable → test it first. Run a one-week experiment where you observe the relevant dynamics before committing to automation. (5) Review the schema periodically — contexts change, and a schema that was accurate last year may not be now. When the schema changes, the agent built on it needs redesign.