prompt-engineer

prompt-engineer comes from the Bernstein project, published by sipyourdrink-ltd and maintained by Alex Chernysh. Bernstein is an open-source orchestration framework for AI agents: you write the policy in code, such as who may do what, what needs approval and what must be recorded, and Bernstein enforces it. prompt-engineer is one of its role templates. It writes and refines instructions for language models, covering system prompt design, few-shot examples, chain-of-thought and structured output, and token budgets. Its rules: state a hypothesis and expected behavior for every prompt change, write evaluation cases alongside it and test against at least three representative inputs, keep prompts in template files, out of application code, comment any model-specific workaround, and only modify the files assigned to the task.

Who it's for:developers building AI applications who need to maintain prompts carefully, especially teams already using Bernstein.

Before you start: used as a role within the Bernstein framework. Bernstein is licensed under Apache 2.0.

Based on the project's public documentation. Not tested by AI Inspiration World.