It is unclear how CRISPR editing outcomes vary across the genome and whether undesirable events such as structural variants (SVs) are predictable or preventable. To define a genome-wide map of editability, we performed whole-genome sequencing on 1875 budding yeast clones edited across 16 chromosomes by CRISPR-Cas9 and donor-templated repair. We found that unintended edits, including short indels and SVs, were enriched in specific genomic and sequence contexts. We developed a predictive model, SCORE (System for CRISPR Outcome and Risk Evaluation), which revealed 4.8% of the genome as SV prone, consisting of 562 SV hotspots. Donor repair-enhancing strategies suppressed SV formation in regions with moderate, but not high, predicted risk. Applying SCORE to the Sc2.0 synthetic yeast genome revealed a markedly altered SV landscape due to the removal of endogenous repetitive elements and the insertion of loxP sites. Our study provides the genome-scale map of SV hotspots after CRISPR editing and predictive and experimental tools to mitigate their formation.