review-iterate

Iterative code review enforces a structured, multi-round workflow where findings are tiered by severity and a stopping rule triggers when only minor polish items remain. This prevents review sessions from turning into unproductive padding by requiring explicit user selection before applying any fixes.

Severity-tiered issue listing groups review findings into red, orange, yellow, and green categories so you can prioritize what matters most for correctness. You explicitly choose which tiered findings to address, and the system fixes only those selected items before re-reviewing the updated scope.

Run a verification pass after implementing changes to catch regressions or newly visible issues. The process requires a fresh re-review of the updated scope that avoids recycling previous findings, ensuring that new problems are detected without endlessly generating extra nitpicks.

Yes, user-driven fix selection is the core mechanism. The system lists issues grouped by severity, provides an explicit recommendation, and then defers to your selection. It applies a fix-only-next-step behavior, ensuring only your chosen items are remediated in the subsequent step.

You should stop iterating when a diminishing-returns threshold is met. The process recommends wrapping up when findings degrade to green polish-only items or when consecutive review rounds produce very little new value, preventing unnecessary repetition.

No, this approach applies specifically to non-security and non-one-shot contexts. It is designed for work-in-progress code or documentation requiring multi-round, user-prioritized verification and iterative refinement rather than immediate or security-focused remediation.