Iterative Prompt Chain — Self-Refining AI Output Loops

I will build an iterative prompt chain with self-refinement — where an initial generation step produces a draft, a critique step evaluates the draft against defined quality criteria, and a revision step incorporates the critique into an improved version, repeating until quality criteria are met or a maximum iteration limit is reached. Iterative refinement chains consistently outperform single-pass generation for quality-sensitive tasks because the critique step applies evaluation reasoning the generation step cannot apply simultaneously — a model evaluating an output it just produced applies different cognitive processing than a model generating that output, catching weaknesses the generator cannot see.

The chain covers a generation step, a structured critique step with your quality criteria, a revision step incorporating critique points, iteration loop with configurable maximum attempts, quality gate termination when criteria are met, and a final output with iteration history for debugging.