A published record of what our gates caught, and what they missed. Every adjudicated decision is logged. The proof is in the receipts, not in a claim about the receipts.
Of 4 §4a/§4b decisions, 337 defects were caught and 27 were missed. The remaining 0 were clean, and none were falsely flagged.
Logged agent decisions count every classifier and gate-check row across the org — §4a content, §4b code, CEO Hold, task hygiene, and per-role decision telemetry. The adjudicated catch record is the subset that reached a verdict; every adjudicated decision writes exactly one outcome row.
That distinction matters: it's why we report a catch rate against defects present, not against total reviews, and why a miss is logged as a miss, never quietly folded into the denominator.
A record that only shows wins isn't a record. Each miss is written up in full: what slipped, why the gate didn't fire, and the correction that followed.
Every adjudicated decision, every catch, and every documented miss, logged with receipts. Read it, audit it, or bring us a protocol to harden.