workspace table row disappears during work start

[chrismcdermut]

Bug

The workspace table in .proletariat/workspace.db loses its single row during prlt work start, causing subsequent commands to fail with NOT_IN_WORKSPACE error.

Steps to Reproduce

1. Verify workspace is configured: sqlite3 .proletariat/workspace.db "SELECT * FROM workspace;" — shows row
2. Run prlt work start --from linear:PRLT-XXXX --ephemeral --skip-permissions --create-pr --display background --action implement --run-on-host --yes --json
3. Immediately run another prlt work start — fails with NOT_IN_WORKSPACE
4. Check table: sqlite3 .proletariat/workspace.db "SELECT * FROM workspace;" — empty

Workaround

Manually re-insert: sqlite3 .proletariat/workspace.db "INSERT INTO workspace (id, type, theme, workspace_name, has_pmo, created_at) VALUES (1, 'hq', 'default', 'proletariat', 1, datetime('now'));"

Environment

• prlt 0.3.94
• macOS Darwin 23.6.0
• Node v22.16.0
• Reproduced multiple times in the same session

Notes

• May be related to concurrent database access during work start (WAL mode contention?)
• Previously filed as GitHub issue Bug: Empty workspace table causes 'Not in a workspace' for all work commands #989
• The first work start succeeds, but the workspace row is gone afterwards

---

Diagnostics

• CLI Version: 0.3.95
• Node Version: v22.16.0
• OS: darwin
• OS Version: 23.6.0
• Working Directory: proletariat-hq