Background: Most Zimbabweans access medical care through tiered health systems, with decentralized HIV care and centralized oncology care. Local health centre staff are trained in ART administration but not in AIDS-associated Kaposi sarcoma (AIDS-KS) diagnosis or management. To improve AIDS-KS prognoses, interventions on screening, diagnosis, and treatment must be integrated into primary HIV care settings. We conducted a community-based randomized step-wedge cluster trial for a package of KS care training interventions. Methods: Interventions intended to improve KS detection and management included a standardized KS clinical evaluation tool, palliative care integration, standardized treatment, and improved consultative services. Interventions were evaluated at eight Zimbabwe primary HIV care sites by monitoring KS diagnoses and outcomes, including early diagnosis (T0 vs T1 tumor stage), patient retention, and mortality. Analyses controlled for within clinic correlations. Findings: 1102 suspected KS patients (96% HIV+) were enrolled: 47% newly diagnosed, 20% previously diagnosed, and 33% determined as not KS (incorrectly diagnosed). Diagnosis at T0 stage increased post-intervention, though not significant statistically (adjusted odds ratio (aOR)=1·48 (95% confidence interval (95%CI):0·66-3·79);p=0·37). New KS diagnosis rates increased 103% (95%CI:11%-273%);p=0·02) post-intervention, with parallel increased odds of incorrect diagnosis (aOR=2·08 (95%CI:0·33-3·24);p=0·001). Post-intervention, non-significant changes in 90-day return rates (adjusted hazard ratio (aHR)=0·69 (95%CI:0·38-1·45);p=0·21) and survival (aHR=1·36 (95%CI:0·85-2·20);p=0·20) were estimated. Interpretation: KS training interventions increased overall and incorrect KS diagnosis, but not early KS diagnosis, 90-day return rates, or survival at HIV primary care sites. Approaches to increase KS diagnosis accuracy in low-resource settings are needed. Trial Registration: Study design details are available at ClinicalTrials.gov NCT01764360. Funding: Funding was provided by the National Cancer Institute, National Institutes of Health: grant 1R01CA172050. Declaration of Interest: None to declare. Ethical Approval: The study was approved by the institutional review boards of Zimbabwe (Medical Research Council of Zimbabwe, MRCZ), University of Colorado Anschutz Medical Campus (COMIRB), and the Joint Parirenyatwa Hospital and University of Zimbabwe College of Health Sciences Research Ethics Committee (JREC).