nx_team_scorecard.nx

nx_team_scorecard.nx -- the TEAMMATE growth loop (operator: consistently grade the TEAMMATES, not the capabilities, as we work, so the team improves WITHOUT REGRESSION and each teammate becomes MORE capable -- a loop Claude is tied into so the operator isn't manually suggesting growth work, run until the team beats the AI benchmarks). Each ORGAN (Engineer, Doctor, Builder, ... Critic, Caretaker) is scored on capability dimensions; the loop: 1. GRADES every teammate (0-100) -> score per round. 2. REGRESSION GUARD: any teammate whose score DROPPED vs last round is an alarm -- the team must only get stronger; a build that weakens another organ is caught. 3. GROWTH TARGET: the below-benchmark teammate with the largest (gap * leverage) -> the NEXT thing Claude builds. Deterministic, not hand-picked. 4. BENCHMARK: when every teammate >= the benchmark, the team has beaten the bar. So Claude is the builder INSIDE the loop: grade -> pick weakest-highest-leverage -> build it -> re-grade (no regression) -> repeat until all pass. license_tier: ORIGINAL Refs: M0-M4 autonomy ladder.