challenge-loop
Challenge Loop — adversarial hardening for judgment-containing outputs. Two modes: inline self-refutation (zero latency) and independent challenger subagent....
