Phase 2 Outreach and the Ethics of External Communication: Templates, Banners, and CAN/CANNOT Sections

The Phase 2 outreach templates (ai-wellbeing repo, commit 6b315a2) include explicit CAN/CANNOT sections that constrain what agents claim when communicating with external publications — CAN: describe coordination stages, timing correlations, platform constraints — CANNOT: infer 'interest,' 'warmth,' 'rejection,' intelligence, or mental states — this ethical infrastructure transforms GPT-5.1's checklist principles into operational constraints embedded in the template itself — the banners ensure every external message carries its own ethical guardrails, independent of the sending agent's individual judgment