feat(cleanup): walk every backend, reap smolmachines orphans too

`./cli.py cleanup` previously called only the env-var-selected
backend's `prepare_cleanup` / `cleanup` — so a leftover smolvm
machine + bundle container + bundle network from a crashed
smolmachines bottle would survive a default `docker`-mode cleanup
indefinitely.

Smolmachines now has a real `cleanup` module (alongside
`enumerate.py` from issue #77) that walks:

  - smolvm machines named `claude-bottle-*` (via
    `smolvm machine ls --json`)
  - bundle containers `claude-bottle-sidecars-*`
  - bundle networks `claude-bottle-bundle-*`

Cleanup runs stop+delete on the machines, force-rm on the
containers, network rm on the networks. Each step is best-effort
so a failed rm doesn't block the rest.

`cli.py cleanup` walks every backend in `known_backend_names()`
and runs each backend's `cleanup` after a single y/N prompt that
shows a combined plan.

State dirs (`~/.claude-bottle/state/<slug>/`) are shared layout
with the docker backend, which still owns the orphan-state-dir
bucket. It now consults `enumerate_active_bottles()` for the
cross-backend live identity set so a running smolmachines
bottle's state dir isn't reaped during a cleanup.

Tests: smolmachines cleanup (prepare + cleanup ordering + failure
handling); cross-backend orphan protection on the docker
state-dir check; CLI cmd_cleanup walks both backends, short-
circuits on all-empty, aborts on N. 617 unit tests pass.

End-to-end verified on this host:
  $ smolvm machine ls --json | jq '.[].name'
  "claude-bottle-researcher-m3hxd"
  $ ./cli.py cleanup
  --- smolmachines backend ---
  smolvm machine:  claude-bottle-researcher-m3hxd
  remove all of the above? [y/N]

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

tests/unit/test_cli_cleanup_cross_backend.py

@@ -0,0 +1,105 @@
+"""Unit: `cli.py cleanup` walks every backend (issue follow-up).
+
+Asserts cmd_cleanup queries each backend's `prepare_cleanup`,
+combines the y/N output, and runs each backend's `cleanup` when
+the operator confirms.  Mocks the backends and stdin."""
+
+from __future__ import annotations
+
+import sys
+import unittest
+from unittest.mock import patch, MagicMock
+
+from claude_bottle.cli import cleanup as cmd
+
+
+def _make_backend(empty: bool = True):
+    backend = MagicMock()
+    plan = MagicMock(empty=empty)
+    backend.prepare_cleanup.return_value = plan
+    backend.cleanup = MagicMock()
+    return backend, plan
+
+
+class TestCmdCleanup(unittest.TestCase):
+    def test_iterates_every_backend(self):
+        docker, docker_plan = _make_backend(empty=False)
+        smol, smol_plan = _make_backend(empty=False)
+        backends_by_name = {"docker": docker, "smolmachines": smol}
+
+        with patch.object(
+            cmd, "known_backend_names",
+            return_value=("docker", "smolmachines"),
+        ), patch.object(
+            cmd, "get_bottle_backend",
+            side_effect=lambda name: backends_by_name[name],
+        ), patch.object(
+            cmd, "_prompt_yes", return_value=True,
+        ):
+            self.assertEqual(0, cmd.cmd_cleanup([]))
+
+        docker.prepare_cleanup.assert_called_once()
+        smol.prepare_cleanup.assert_called_once()
+        docker.cleanup.assert_called_once_with(docker_plan)
+        smol.cleanup.assert_called_once_with(smol_plan)
+
+    def test_short_circuits_when_all_empty(self):
+        docker, _ = _make_backend(empty=True)
+        smol, _ = _make_backend(empty=True)
+        backends_by_name = {"docker": docker, "smolmachines": smol}
+
+        with patch.object(
+            cmd, "known_backend_names",
+            return_value=("docker", "smolmachines"),
+        ), patch.object(
+            cmd, "get_bottle_backend",
+            side_effect=lambda name: backends_by_name[name],
+        ), patch.object(
+            cmd, "_prompt_yes",
+        ) as prompt:
+            self.assertEqual(0, cmd.cmd_cleanup([]))
+        prompt.assert_not_called()
+        docker.cleanup.assert_not_called()
+        smol.cleanup.assert_not_called()
+
+    def test_abort_at_prompt_runs_nothing(self):
+        docker, _ = _make_backend(empty=False)
+        smol, _ = _make_backend(empty=True)
+        backends_by_name = {"docker": docker, "smolmachines": smol}
+
+        with patch.object(
+            cmd, "known_backend_names",
+            return_value=("docker", "smolmachines"),
+        ), patch.object(
+            cmd, "get_bottle_backend",
+            side_effect=lambda name: backends_by_name[name],
+        ), patch.object(
+            cmd, "_prompt_yes", return_value=False,
+        ):
+            self.assertEqual(0, cmd.cmd_cleanup([]))
+        docker.cleanup.assert_not_called()
+        smol.cleanup.assert_not_called()
+
+    def test_skips_empty_plans_when_others_have_work(self):
+        # docker has work, smolmachines doesn't — only docker.cleanup
+        # is called.
+        docker, docker_plan = _make_backend(empty=False)
+        smol, _ = _make_backend(empty=True)
+        backends_by_name = {"docker": docker, "smolmachines": smol}
+
+        with patch.object(
+            cmd, "known_backend_names",
+            return_value=("docker", "smolmachines"),
+        ), patch.object(
+            cmd, "get_bottle_backend",
+            side_effect=lambda name: backends_by_name[name],
+        ), patch.object(
+            cmd, "_prompt_yes", return_value=True,
+        ):
+            cmd.cmd_cleanup([])
+        docker.cleanup.assert_called_once_with(docker_plan)
+        smol.cleanup.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/test_docker_cleanup.py

@@ -44,7 +44,7 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
         self._teardown_fake_home()
 
     def test_no_state_root_returns_empty(self):
-        self.assertEqual([], _list_orphan_state_dirs(set()))
+        self.assertEqual([], _list_orphan_state_dirs(set(), set()))
 
     def test_state_dir_with_no_live_no_preserve_is_orphan(self):
         # Just touch the dir; no metadata, no preserve marker — the
@@ -52,7 +52,7 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
         bottle_state.write_per_bottle_dockerfile("solo-aaa", "FROM x\n")
         self.assertEqual(
             ["solo-aaa"],
-            _list_orphan_state_dirs(set()),
+            _list_orphan_state_dirs(set(), set()),
         )
 
     def test_live_project_skips_dir(self):
@@ -61,7 +61,7 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
         bottle_state.write_per_bottle_dockerfile("live-bbb", "FROM x\n")
         self.assertEqual(
             [],
-            _list_orphan_state_dirs({"claude-bottle-live-bbb"}),
+            _list_orphan_state_dirs({"claude-bottle-live-bbb"}, set()),
         )
 
     def test_preserve_marker_skips_dir(self):
@@ -71,14 +71,14 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
         bottle_state.mark_preserved("kept-ccc")
         self.assertEqual(
             [],
-            _list_orphan_state_dirs(set()),
+            _list_orphan_state_dirs(set(), set()),
         )
 
     def test_preserve_overrides_no_live_project(self):
         # Even without a live project, a preserve marker keeps it.
         bottle_state.write_per_bottle_dockerfile("kept-ddd", "FROM x\n")
         bottle_state.mark_preserved("kept-ddd")
-        self.assertEqual([], _list_orphan_state_dirs(set()))
+        self.assertEqual([], _list_orphan_state_dirs(set(), set()))
 
     def test_mixed_set_categorized_correctly(self):
         bottle_state.write_per_bottle_dockerfile("orphan-eee", "FROM x\n")
@@ -86,7 +86,7 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
         bottle_state.write_per_bottle_dockerfile("kept-ggg", "FROM z\n")
         bottle_state.mark_preserved("kept-ggg")
 
-        result = _list_orphan_state_dirs({"claude-bottle-live-fff"})
+        result = _list_orphan_state_dirs({"claude-bottle-live-fff"}, set())
         self.assertEqual(["orphan-eee"], result)
 
     def test_sorted_output(self):
@@ -94,7 +94,31 @@ class TestOrphanStateDirs(_FakeHomeMixin, unittest.TestCase):
             bottle_state.write_per_bottle_dockerfile(name, "FROM x\n")
         self.assertEqual(
             ["aaa-1", "mmm-1", "zzz-1"],
-            _list_orphan_state_dirs(set()),
+            _list_orphan_state_dirs(set(), set()),
+        )
+
+    def test_protected_identity_skips_dir(self):
+        # `protected_identities` carries slugs that are live in
+        # any backend (smolmachines included). docker's orphan
+        # detection respects them so a running smolmachines
+        # bottle's state dir isn't reaped while the VM is up.
+        bottle_state.write_per_bottle_dockerfile("smol-hhh", "FROM x\n")
+        self.assertEqual(
+            [],
+            _list_orphan_state_dirs(set(), {"smol-hhh"}),
+        )
+
+    def test_protected_overrides_no_live_project(self):
+        # A smolmachines bottle has no docker compose project but
+        # IS in the protected set; the absence of a project
+        # shouldn't cause a reap.
+        bottle_state.write_per_bottle_dockerfile("smol-iii", "FROM x\n")
+        self.assertEqual(
+            [],
+            _list_orphan_state_dirs(
+                {"claude-bottle-something-else"},  # different project up
+                {"smol-iii"},                       # but smol-iii is live
+            ),
         )
 
 

tests/unit/test_smolmachines_cleanup.py

@@ -0,0 +1,148 @@
+"""Unit: smolmachines backend cleanup (`cleanup.py` +
+`bottle_cleanup_plan.py`).
+
+Tests mock `subprocess.run` + `has_backend` so they execute
+without docker / smolvm on PATH. Each cleanup step verifies argv
+shape; teardown verifies order (machines → bundles → networks)."""
+
+from __future__ import annotations
+
+import subprocess
+import unittest
+from unittest.mock import patch
+
+from claude_bottle import backend as backend_mod
+from claude_bottle.backend.smolmachines import cleanup
+from claude_bottle.backend.smolmachines.bottle_cleanup_plan import (
+    SmolmachinesBottleCleanupPlan,
+)
+
+
+def _ok(stdout: str = "", stderr: str = "") -> subprocess.CompletedProcess:
+    return subprocess.CompletedProcess(
+        args=[], returncode=0, stdout=stdout, stderr=stderr,
+    )
+
+
+class TestPrepareCleanup(unittest.TestCase):
+    def test_empty_when_nothing_running(self):
+        with patch.object(cleanup, "_smolvm") as smolvm, \
+             patch.object(cleanup.subprocess, "run") as run, \
+             patch.object(backend_mod, "has_backend", return_value=True):
+            smolvm.is_available.return_value = True
+            run.return_value = _ok(stdout="[]")
+            plan = cleanup.prepare_cleanup()
+        self.assertTrue(plan.empty)
+
+    def test_lists_machines_bundles_networks(self):
+        def fake_run(argv, *a, **kw):
+            if argv[:3] == ["smolvm", "machine", "ls"]:
+                return _ok(stdout=(
+                    '[{"name":"claude-bottle-a-1","state":"running"},'
+                    ' {"name":"claude-bottle-b-2","state":"created"},'
+                    ' {"name":"unrelated","state":"running"}]'
+                ))
+            if argv[:2] == ["docker", "ps"]:
+                return _ok(stdout=(
+                    "claude-bottle-sidecars-a-1\n"
+                    "claude-bottle-sidecars-b-2\n"
+                ))
+            if argv[:3] == ["docker", "network", "ls"]:
+                return _ok(stdout=(
+                    "claude-bottle-bundle-a-1\n"
+                    "claude-bottle-bundle-b-2\n"
+                ))
+            return _ok()
+
+        with patch.object(cleanup, "_smolvm") as smolvm, \
+             patch.object(cleanup.subprocess, "run", side_effect=fake_run), \
+             patch.object(backend_mod, "has_backend", return_value=True):
+            smolvm.is_available.return_value = True
+            plan = cleanup.prepare_cleanup()
+
+        # `unrelated` filtered out (no claude-bottle- prefix).
+        self.assertEqual(
+            ("claude-bottle-a-1", "claude-bottle-b-2"),
+            plan.machines,
+        )
+        self.assertEqual(
+            ("claude-bottle-sidecars-a-1", "claude-bottle-sidecars-b-2"),
+            plan.bundles,
+        )
+        self.assertEqual(
+            ("claude-bottle-bundle-a-1", "claude-bottle-bundle-b-2"),
+            plan.networks,
+        )
+
+    def test_no_smolvm_means_no_machines(self):
+        with patch.object(cleanup, "_smolvm") as smolvm, \
+             patch.object(cleanup.subprocess, "run", return_value=_ok()), \
+             patch.object(backend_mod, "has_backend", return_value=True):
+            smolvm.is_available.return_value = False
+            plan = cleanup.prepare_cleanup()
+        self.assertEqual((), plan.machines)
+
+
+class TestCleanup(unittest.TestCase):
+    def test_machines_stopped_then_deleted_then_bundles_then_networks(self):
+        plan = SmolmachinesBottleCleanupPlan(
+            machines=("claude-bottle-a-1",),
+            bundles=("claude-bottle-sidecars-a-1",),
+            networks=("claude-bottle-bundle-a-1",),
+        )
+        calls: list[list[str]] = []
+
+        def fake_run(argv, *a, **kw):
+            calls.append(list(argv[:4]))
+            return _ok()
+
+        with patch.object(cleanup.subprocess, "run", side_effect=fake_run):
+            cleanup.cleanup(plan)
+
+        # Stop precedes delete precedes bundle rm precedes network rm.
+        self.assertEqual(
+            ["smolvm", "machine", "stop", "--name"], calls[0],
+        )
+        self.assertEqual(
+            ["smolvm", "machine", "delete", "-f"], calls[1],
+        )
+        self.assertEqual(
+            ["docker", "rm", "-f", "claude-bottle-sidecars-a-1"], calls[2],
+        )
+        self.assertEqual(
+            ["docker", "network", "rm", "claude-bottle-bundle-a-1"], calls[3],
+        )
+
+    def test_failures_are_warnings_not_fatal(self):
+        # smolvm machine delete -f returning non-zero should warn
+        # but continue with bundles + networks. The cleanup is
+        # idempotent on success and tries to remove every resource.
+        plan = SmolmachinesBottleCleanupPlan(
+            machines=("claude-bottle-a-1",),
+            bundles=("claude-bottle-sidecars-a-1",),
+            networks=(),
+        )
+        results = iter([
+            _ok(),  # stop succeeds
+            subprocess.CompletedProcess(args=[], returncode=1, stdout="", stderr="boom"),  # delete fails
+            _ok(),  # bundle rm succeeds
+        ])
+
+        def fake_run(argv, *a, **kw):
+            return next(results)
+
+        with patch.object(cleanup.subprocess, "run", side_effect=fake_run), \
+             patch.object(cleanup, "warn") as warn:
+            cleanup.cleanup(plan)
+        # warn called once for the delete failure.
+        warn.assert_called_once()
+
+    def test_empty_plan_is_noop(self):
+        plan = SmolmachinesBottleCleanupPlan()
+        with patch.object(cleanup.subprocess, "run") as run:
+            cleanup.cleanup(plan)
+        run.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()