Despite DNA nanotechnology has spawned a broad variety and taken a giant leap toward cancer theranostic applications over the last decade, the homogeneous DNA nanostructures often suffer from fatal degradation due to their limited stability and specificity. Herein, for the first time, we report a stable DNA tetrahedra–gold nanoclusters (DT/AuNCs) nanohybrid with a self-assembly/programmed disassembly manner for stimuli-responsive tumor imaging and gene-chemo therapy. By utilizing the multifunctional peptides with positive and legumain-specific domains as bioligands, AuNCs were synthesized as signal generators and gate guard attached on the dual-responsive DT, forming the DT/AuNCs with sequential response to legumain-TK1 mRNA & glutathione. The tumorous biomarker of legumain initiated the signal generation relying on the nanosurface energy transfer effect of AuNCs and denudation of DT-Dox (preliminary disassembly). Successively, the dual-responsive DT-Dox administrated a sequential fragmentation along with Dox release in response to the up-regulated glutathione and TK1 mRNA (secondary disassembly), thereby leading to combined gene silencing and chemo-therapy. The results revealed that the DT/AuNCs nanohybrids significantly improved the stability and enhanced the therapeutic efficiency compared to naked DT. Endowing with remarkable stability against biological milieu and site specificity for drug release, our work exhibits a new prospect of fabricating DNA-based nanohybrids for precise tumor theranostics.