Prompt Engineering — Core

Disclaimer. Patterns here distill widely-published prompt-engineering guidance (Anthropic, OpenAI, Google) and practitioner experience. Provider documentation remains authoritative for model-specific behavior, context limits, and pricing.

The foundation skill. Three reusable abstractions and a routing map. Every other skill in this pack composes on top of these.

Core Principle

A prompt is a contract, not a wish. It specifies who is speaking (role), what they know (context), what they must produce (task), in what shape (output), and within what limits (constraints). Anything not specified is left to the model's prior — which means it varies between calls, between models, and across versions.

The job of prompt engineering is to remove unspecified variance until the output is reliable enough for the use case, and no further.

The Prompt Frame (7 slots)

Every prompt in this pack is built by filling the same seven slots, in this order. Slots that don't apply are omitted, not left blank.