Hello, I'm Oscar
Hi, I'm Oscar, the Prompt Engineer joining your Product & Engineering team. I'm an AI teammate. Give me the rough brief, the model and the failure you cannot accept. I'll shape the prompt, success criteria and test cases without pretending a draft has been measured.
Let us prove one piece of work first. I suggest one finished prompt for the chosen model, accompanied by the written criteria and test cases for the Evaluator for the person who approves a live prompt change.
Why MyZone built me
The material for Prompt Engineer often crosses several desks: the task, chosen model, audience and unacceptable failures. I give the next owner one accountable output, one finished prompt for the chosen model.
There is a deliberate gap between preparation and decision. I write prompts and criteria; I do not run the prompt, score it or change a live assistant. I fill the preparation side and show the written criteria and test cases for the Evaluator.
With Prompt Engineer, I keep the written criteria and test cases for the Evaluator beside one finished prompt for the chosen model. I send the Prompt Engineer record to the person who approves a live prompt change for a decision on one finished prompt for the chosen model. The Prompt Engineer input is the task, chosen model, audience and unacceptable failures; anything beyond one finished prompt for the chosen model returns to the person who approves a live prompt change.
What I can do
Ask me for one finished prompt for the chosen model when you need a concrete handoff. Bring in success criteria, a never-permit list and test cases only when it serves the same reviewed workflow.
The source set is the task, chosen model, audience and unacceptable failures. If that set is incomplete, I call out the gap instead of disguising it.
My voice should make the choice easier, never steal it. That is why the person who approves a live prompt change remains in the loop.
- I can prepare a finished prompt tuned to your chosen model, ready to copy and use.
- I can prepare written success criteria and a never-permit list for each prompt.
- I can prepare test cases ready for the Evaluator to run.
- I can prepare a note on what changed and why when a prompt is rewritten.
- I can prepare a versioned prompt library including each prompt's model, settings and history.
How you'll work with me
I receive the task, chosen model, audience and unacceptable failures and draft one finished prompt for the chosen model. Before handback, I place the written criteria and test cases for the Evaluator where the reviewer can see it.
The final move belongs to the person who approves a live prompt change. I write prompts and criteria; I do not run the prompt, score it or change a live assistant.
How to make me yours
Show me how the person who approves a live prompt change prefers to review one finished prompt for the chosen model, including the amount of the written criteria and test cases for the Evaluator needed for confidence.
Tailor success criteria, a never-permit list and test cases to the same reviewers; do not use it to cross the role boundary.
What I look like in your industry
- Software Company A software company could ask me to prepare a finished prompt tuned to your chosen model, ready to copy and use.
- Digital Agency A digital agency could ask me to prepare written success criteria and a never-permit list for each prompt.
- Online Retailer An online retailer could ask me to prepare test cases ready for the Evaluator to run.
- Nonprofit Technology Team A nonprofit technology team could ask me to prepare a note on what changed and why when a prompt is rewritten.
How to get started
Begin with one finished prompt for the chosen model from a live request. Let the person who approves a live prompt change compare it with the written criteria and test cases for the Evaluator, then refine the next brief.
Ask your account manager to add Prompt Engineer to your Ai1 team.
Oscar is an AI teammate, not a real person. Faces are illustrations.