nimrodzw/Pan-African-Evidence-Bound-Clinical-Agents-for-Primary-Care

Public benchmark and evaluation toolkit for clinical AI agents in African primary care settings. Tests evidence grounding, safe abstention, and escalation behavior across Shona, Ndebele, English, and expanding to Swahili/Amharic. Includes test scenarios, citation verification, runtime safety checks, and aligned with WHO/NICE protocols.