panel-review-loop

Runs the review-fix-rereview cycle around the panel-review skill until the work is in good shape, then reports back. You do the task, fan a panel review out to independent agents, decide which findings are genuinely worth fixing, fix those, and review again — looping until nothing worth fixing is left. The deliverable is a short report: what you fixed, what you chose to leave alone (and why), and anywhere the implementation drifted from the goal.

This is autonomous by design: once started, it runs the whole loop without stopping for approval between rounds, and reports once at the end. You'll still see panel-review's own per-panelist progress stream each round — that's progress output, not a check-in; the loop simply never pauses for your input until it's done.

When to use