Current gene-editing approaches for treatment of recessive dystrophic epidermolysis bullosa (RDEB), an inherited, severe form of blistering skin disease, suffer from low efficiencies and safety concerns that complicate implementation in clinical settings. We present a strategy for efficient and precise repair of RDEB-associated mutations in the COL7A1 gene. We compared the efficacy of double-strand breaks (induced by CRISPR/Cas9), single nicks, or double nicks (induced by Cas9n) in mediating repair of a COL7A1 splice-site mutation in exon 3 by homologous recombination (HR). We accomplished remarkably high HR frequencies of 89% with double nicking while at the same time keeping unwanted repair outcomes, such as non-homologous end joining (NHEJ), at a minimum (11%). We also investigated the effects of subtle differences in repair template design on HR rates and found that strategic template-nicking can enhance COL7A1-editing efficiency. In RDEB patient keratinocytes, application of double-nicking led to restoration and subsequent secretion of type VII collagen at high efficiency. Comprehensive analysis of 25 putative off-target sites revealed no off-target activity for double-nicking, while usage of Cas9 resulted in 54% modified alleles at one site. Taken together, our work provides a framework for efficient, precise, and safe repair of COL7A1, which lies at the heart of a future curative therapy of RDEB.