feature-image-iterate

You can automate feature image iteration by draining pending workflow items from the dev API, selecting a strategy per item, and executing payloads to generate, recomposite, or clarify images based on user feedback.

The best way to refine AI art assets in a feedback loop is to process each pending image iteration item individually, apply user feedback to select a specific strategy, and update the workflow status after generating the response.

Select between generating, recompositing, or clarifying an image by evaluating user feedback and snapshot context for each pending workflow item, ensuring exactly one strategy is applied per item to construct the execution payload.

Drain pending feature image workflow items in claude-code by executing the drain script via tsx to load context, then running the respond script with a prepared payload to apply the chosen action and post the assistant message.

The feature image iteration workflow does not require external dependencies, operating entirely within the claude-code environment using internal scripts to drain pending items and construct execution payloads.

Feature image iteration items remain pending when the workflow has not been fully drained and processed, requiring each item to receive a selected strategy and a corresponding payload execution before marking the workflow as applied.