aif-loop

Automate iterative refinement by running a structured multi-phase loop that applies strict phase contracts and persistent state across sessions. This drives artifact evolution through PLAN, PRODUCE, PREPARE, EVALUATE, CRITIQUE, and REFINE stages until quality gates pass.

A phase contract in task automation enforces strict, reproducible execution rules for each stage of an iteration loop. It ensures deterministic artifact improvement by loading thresholds and validation rules from references before evaluating or refining outputs.

Resume artifact generation loops by relying on persistent state that tracks current progress, run details, and history. The workflow saves these states alongside artifact.md across sessions, enabling seamless auditing and continuation of iterative improvement tasks.

Yes, iterative improvement workflows support parallel execution for PRODUCE and PREPARE phases alongside EVALUATE. This parallelization accelerates artifact evolution while maintaining strict phase contracts and reproducible scoring across iterations.

You need quality gates for iterative content generation when refining API specs, docs, or code artifacts requires passing specific scoring thresholds. The workflow enforces these gates by loading rules from references and looping until criteria are met or iteration limits are reached.

No, aif-loop operates without external dependencies. It uses internal references to load thresholds and rules, requiring only a task prompt to initialize the six-phase loop and begin guided artifact refinement.