recognizable-quality-test-generator

To create an LLM-as-judge prompt, provide a workflow description and 3–5 historical outputs. The skill generates a self-contained judge prompt with verifiable criteria, pass/fail conditions, and calibration steps to evaluate agent outputs.

A quality rubric for agent evaluation is a structured set of 4–8 verifiable criteria used to judge outputs. You need one when an agent requires formal pass/fail success criteria and self-evaluation steps to ensure output reliability.

You calibrate evaluation criteria by inputting 3–5 historical outputs into the generator. It applies a calibration workflow that refines the rubric based on those outcome examples, producing a formal evaluation standard.

Yes, you can embed the generated judge rubric directly. The skill outputs a self-contained LLM-as-judge prompt that includes workflow context, criteria, final verdict logic, and pass/fail conditions for integration into agent evaluation steps.

You need a workflow description and 3–5 historical outputs to generate a structured quality rubric. These inputs allow the tool to produce formal evaluation criteria and a self-contained judge prompt for prompt engineering workflows.