Recover archive failures and retained work

Scope

The selected task only. Archive stops owned processes/services and preserves data. Deleting a checkout and deleting its volumes are separate operations.

Diagnose

Use Orca desktop Delete for the verified SSH hook path. On Orca 1.4.220 CLI worktree rm --run-hooks refuses remote hook execution; do not interpret that as successful cleanup. Read the hook failure before changing files.

On Shipyard inspect the task’s lifecycle status, private archive log, tmux session and project-owned containers. Read the active dashboard ownership record before stopping the shared preview. Do not guess volume names from a directory basename.

Recovery and preservation

If the task is authorized to stop, run its explicit archive command and verify owned processes/containers stopped while unrelated tasks remain healthy. For API, companion dashboard edits are retained under ~/dev-environment/state/orca/retained/<task-id>/dashboard. Use Git status there before any further cleanup. Re-enabling the same task can move its companion back when the API tree exists.

Appliance state stays in the common Git directory and named volumes. Worktree identity depends on its absolute path; do not rename active worktrees or assume a new path will attach old state.

Verification and escalation

Confirm retained files, tracked/untracked edits and volumes before removing anything else. A CLI archive waiver may only be considered after successful explicit cleanup and within authorized deletion scope. If ownership is ambiguous or a process will not stop gracefully, preserve state and investigate. No global Docker prune, forced kill or automatic volume deletion.