Prompt Engineer as Playwright

Treat the model-facing transcript as a script you author — roles, injected turns, and tools — distinct from the human’s visible conversation.

In chat apps, the human↔assistant dialogue is not identical to the application↔model transcript. The harness may fabricate supporting user or assistant turns (highlighted code the user never typed, retrieved docs) before the model replies. Berryman’s metaphor:

• Characters — ChatML roles (system, user, assistant, tool).
• Script — the assembled prompt working the user’s problem.
• Playwrights — prompt engineer (structure/boilerplate), human user (theme), LLM (assistant lines), and external APIs (injected content).

The prompt engineer is lead playwright / showrunner: shape tone and information flow so the model sees a coherent document under system prompt architecture and conversational agent context. Never confuse “what the user said” with “what the model must see.”

---

Sources

• Prompt Engineering for LLMs (Berryman), ch. 3