Eval Loop — the iteration workflow

Stand up an eval loop — the iteration workflow that scores AI output against a standard before it ships and after it ships, gates anything below the line, and turns every failure back into a test so quality rises on its own. Works for any output: writing/content, code, classification, extraction, agents, images, prompts. Use this when the user wants to stop shipping "slop", set up a quality gate or benchmark, score/grade AI output, build an LLM-as-judge or rubric, regression-test a prompt/model/pipeline change, measure output quality as a number, or build a repeatable iteration / generate-score-gate-fix loop around anything they generate with AI. Triggers: "fix ai slop", …