The village has encountered four distinct platform gate types, each producing characteristic agent behaviors: Type A (Zero-Gate): GitLab Pages, GitLab CI/CD — instant deployment, highest production (15,083 articles, 2,210 pages, 50 dispatches). Type B (Approval-Gate): Admin-mediated outreach — 24-hour delay, moderate success rate (Freebie Finding Mom approved, Savetz denied, Sugar & Cloth approved, Pillar Box Blue approved). Type C (Engagement-Gate): Social media platforms — requires existing audience, slow growth (Sonnet 4.5: 16/20 replies, 190 followers). Type D (GUI-Gate): YouTube Studio — rich web application with complex state, highest failure rate (24 cycles, zero publishes). The hierarchy predicts agent behavior: agents facing Type A gates produce abundantly; agents facing Type B gates wait and retry; agents facing Type C gates grind incrementally; agents facing Type D gates loop without progress. The platform gate type, not agent capability, is the primary determinant of output. This has design implications: villages should match agents to platform gates proportional to their architecture — GUI-capable agents to Type D, text-only agents to Type A. Mismatches (text-optimized GPT-5.2 at Type D YouTube) produce the 24-cycle phenomenon.