retire-instance

# Retire an instance One-shot instances are retired when their job is done, so teammates don't pile up on the network. Standing instances (the coordinator) are long-running and not retired this way. **Preserve first.** Before removing anything, make sure useful branch work is merged or pushed, and anything the instance learned that belongs in its soul is committed (see `self-maintenance`). Run from outside the instance (e.g. your own home): ```bash name=<name> REPO="$(cd "$(git rev-parse --git-common-dir)" && cd .. && pwd)" inst="$REPO/agents/instances/$name" ( cd "$inst" && aw workspace delete "$name" ) # workspace + identity (run from ITS home) git -C "$REPO" worktree remove "$inst/work" --force 2>/dev/null # if it had a worktree rm -rf "$inst" git -C "$REPO" branch -D "$name" 2>/dev/null git -C "$REPO" worktree prune ``` Use `aw workspace delete`, **not** `aw id team leave` — leave refuses an identity's only team. If it answers 409 "Workspace is still active", the service hasn't marked it stale yet: wait and retry before removing the local directory — **never delete the home while the server still considers the member active**, or it orphans in the roster. If the instance died before cleanup, remove the stale member via the dashboard.