Own a11y and i18n
You're leading a small team using AI agents to rebuild a legacy web app's UI, and the product must ship accessible (WCAG AA) and localized into 8 languages from day one. Junior engineers are generating components fast but the output keeps regressing on a11y and i18n. How do you decompose this work for AI agents so these cross-cutting concerns are owned rather than sprinkled, and where do you insist on human checkpoints?
evaluate_merge_gate(component_records: list[str]) → list[str][["name=LoginDialog;axe=2;raw=0;overflow=0;primitive=true"]]out["LoginDialog: block, 2 axe violation(s)"][["name=SettingsRow;axe=0;raw=3;overflow=1;primitive=false","name=Button;axe=0;raw=0;overflow=0;primitive=true"]]out["SettingsRow: block, 3 hardcoded string(s)","Button: human, shared primitive or key schema change"][["name=ProfileCard;axe=0;raw=0;overflow=0;primitive=false"]]out["ProfileCard: auto, composed from primitives with gates green"]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.