What problem does it solve? When multiple parallel sessions append rows to one shared ledger file (a tracker data.js, CHANGELOG, registry, or manifest), a branch whose tree is stale can silently delete other sessions' rows on merge — with no file deletion, no merge conflict, an ordinary-looking diffstat, and passing validators. File-level deletion gates (git diff --diff-filter=D) never fire, and id-only checks miss rows whose bodies were rolled back while their ids survived. ## Core Features & Use Cases - Row-level audit against current origin/main: Parse both sides of the ledger and compare entries by id AND body, reporting REMOVED rows and CHANGED BUT NOT MINE rows that an id-only check cannot see. - Rebuild-based recovery: Take main's ledger wholesale and replay your own scripted edits onto it, rather than hand-resolving interleaved JSON conflicts — additive by construction so it cannot revert anyone. - Post-merge verification: Re-read main after merging to confirm your rows are present and current, and re-audit because the staleness window reopens with every sibling merge. - Use Case: Before merging a long-open PR that edits a shared data.js tracker, run the audit script against origin/main; it names the 3 tasks, 2 decisions, and 8 artifacts your branch would have deleted, then you rebuild by re-running your splicing script on main's file. ## Quick Start Ask the AI to audit whether your current branch would drop or roll back any rows in the shared ledger file compared to origin/main before merging.
