diff --git a/docs/REPOSITORY_SANITATION.md b/docs/REPOSITORY_SANITATION.md new file mode 100644 index 00000000..ef07657b --- /dev/null +++ b/docs/REPOSITORY_SANITATION.md @@ -0,0 +1,16 @@ +# Repository sanitation status + +The current-tree hygiene campaign removed committed local AI/import artifacts from `.codex-input`, including SQLite databases and WAL/SHM state, archives, generated JSON, geometry diagnostics and large contact-sheet screenshots. A deny-by-default `.codex-input/.gitignore` prevents recurrence. + +## HISTORY_REWRITE_REQUIRED + +Deleting these files from the current tree does not remove their objects from reachable Git history. A later owner-approved history phase must inventory all refs and remove or explicitly retain the historical objects after evidence and backup review. + +| Path family | Information type | Why history sanitation is required | +|---|---|---| +| `.codex-input/*.db`, `*.db-wal`, `*.db-shm` | local application/database state | may contain records and internal processing state; WAL/SHM files are runtime artifacts | +| `.codex-input/*.tar`, `*.tar.gz` | source/workspace snapshots | duplicate source and potentially private repository content | +| `.codex-input/*contact_sheet*.png` | generated screenshots/imagery | may expose source imagery, UI state or sensitive locations | +| `.codex-input/*.json`, geometry diagnostics | generated analysis/evidence | may contain private inputs, coordinates or operational context | + +Before any public review, perform a bare-clone all-ref secret, path, database/archive, media and large-object scan. No history was rewritten during this campaign.