Before　herd　immunity　against　Coronavirus　disease　2019　(COVID-19)　is　achieved　by　mass　vaccination,　science-based　guidelines　for　non-pharmaceutical　interventions　are　urgently　needed　to　reopen　megacities.　This　study　integrated　massive　mobile　phone　tracking　records,　census　data　and　building　characteristics　into　a　spatially　explicit　agent-based　model　to　simulate　COVID-19　spread　among　11.2　million　individuals　living　in　Shenzhen　City,　China.　After　validation　by　local　epidemiological　observations,　the　model　was　used　to　assess　the　probability　of　COVID-19　resurgence　if　sporadic　cases　occurred　in　a　fully　reopened　city.　Combined　scenarios　of　three　critical　non-pharmaceutical　interventions　(contact　tracing,　mask　wearing　and　prompt　testing)　were　assessed　at　various　levels　of　public　compliance.　Our　results　show　a　greater　than　50%　chance　of　disease　resurgence　if　the　city　reopened　without　contact　tracing.　However,　tracing　household　contacts,　in　combination　with　mandatory　mask　use　and　prompt　testing,　could　suppress　the　probability　of　resurgence　under　5%　within　four　weeks.　If　household　contact　tracing　could　be　expanded　to　work/class　group　members,　the　COVID　resurgence　could　be　avoided　if　80%　of　the　population　wear　facemasks　and　40%　comply　with　prompt　testing.　Our　assessment,　including　modelling　for　different　scenarios,　helps　public　health　practitioners　tailor　interventions　within　Shenzhen　City　and　other　world　megacities　under　a　variety　of　suppression　timelines,　risk　tolerance,　healthcare　capacity　and　public　compliance.