crabbuild/trail/trail-recovery

Trail Recovery

Treat Trail's blockers as evidence. Diagnose first, preserve user work and durable history, then choose the narrowest recovery that addresses the proven cause.

Start Read-Only

trail --format json status
trail doctor
trail fsck

For a managed task, start with trail agent diagnose <task>, delta --patch, and checkpoints. For a lane, start with lane status, timeline, diff --patch, and readiness.

Read recovery-playbook.md before any rewind, conflict resolution, forced sync, restore, non-dry-run garbage collection, lane removal, or schema recovery.

Never Bypass the Cause

Do not reach for --force, --allow-stale, --allow-ignored, --direct, or --no-auth merely to make a command succeed. Do not delete locks, database files, refs, tokens, journals, quarantine paths, or .trail internals manually.

Back up before invasive recovery. Preview when available. After recovery, rerun the original diagnostic plus status/diff/readiness and report what evidence changed, what remains blocked, and whether any gate was skipped.