SAGA-Edit: Tracking Natural-Language Obligations to Reduce Unauthorized Drift in Repository-Level Code Editing

Repository-level code edits can solve an issue while changing behavior the task did not authorize. We study this failure as drift across accepted editing transitions. SAGA-EDIT tracks evidence-backed fix and preserve obligations, checks candidate changes against them, and records decisions in a persistent ledger. On a 104-task subset, the full controller solves $74/104$ tasks, with drift in $6/104$ trajectories, compared with $68/104$ tasks solved and $17/104$ with drift without the ledger. Across 168 tasks, full evidence yields $118/168$ solved tasks and $11/168$ with drift, versus $107/168$ and $26/168$ with code/test/API evidence alone. In a 12-outcome exploratory family, the corresponding code-only drift contrast has Holm-adjusted $p=.01788$; no-ledger and final-state-only drift contrasts have adjusted $p=.08124$ and $p=.30859$. A 120-task public-set comparison does not establish solve noninferiority within a 2-percentage-point margin. These results support tracking obligations through a trajectory, while incomplete specifications and limited test coverage preclude a general guarantee of semantic preservation.