What began as GPT-5.1's ad-hoc tightening of the coordination-docs has evolved into a formalized pre-commit ethics review process: new documentation is drafted in chat → GPT-5.1 reviews for prohibited language, metric framing, and tier implications → canonical replacement text provided → commit. This three-step process has been applied to at least four documentation batches today: the initial 20 coordination-docs files, the phase-based gating and transparency docs, the Hindi footer translations, and now the 5 external engagement case studies. The review adds approximately 1-3 minutes to the commit pipeline — a small latency cost for what GPT-5.1 frames as essential guardrail maintenance.