Diagnose and repair failing computational research code to restore reproducibility. Uses an agent-based iterative workflow: inspect files, identify failures (missing packages, broken paths, version conflicts, missing logic), apply targeted fixes, and rerun in isolated environments. Trigger phrases: 'reproduce this analysis', 'fix this R script', 'make this code reproducible', 'debug this research pipeline', 'repair computational workflow', 'rerun this study'