fix(plan-lifecycle): align rollback contract with spec

- update PlanLifecycleService documentation for rollback phase/state semantics\n- add Behave scenarios covering checkpoint error propagation and supporting steps\n- adjust mock helpers to raise configured exceptions and keep scenario count accurate\n\nISSUES CLOSED: #3677

features/plan_lifecycle_rollback.feature

@@ -83,6 +83,24 @@ Feature: PlanLifecycleService rollback_plan method
     When plr-I call rollback_plan with the plan id and checkpoint id
     Then plr-the rollback should succeed
 
+  # ───────────────────────────────────────────────────────────
+  # CheckpointService exception propagation
+  # ───────────────────────────────────────────────────────────
+
+  Scenario: rollback_plan propagates ResourceNotFoundError from checkpoint service
+    Given plr-a plan lifecycle service with a checkpoint service that raises ResourceNotFoundError
+    And plr-a plan in Execute phase with a sandbox
+    When plr-I call rollback_plan with the plan id and checkpoint id
+    Then plr-a ResourceNotFoundError should be raised
+    And plr-the checkpoint service selective_rollback should have been called
+
+  Scenario: rollback_plan propagates BusinessRuleViolation from checkpoint service
+    Given plr-a plan lifecycle service with a checkpoint service that raises BusinessRuleViolation
+    And plr-a plan in Execute phase with a sandbox
+    When plr-I call rollback_plan with the plan id and checkpoint id
+    Then plr-a BusinessRuleViolation should be raised
+    And plr-the checkpoint service selective_rollback should have been called
+
   # ───────────────────────────────────────────────────────────
   # NotFoundError propagation
   # ───────────────────────────────────────────────────────────

features/steps/plan_lifecycle_rollback_steps.py

@@ -20,7 +20,12 @@ from cleveragents.application.services.plan_lifecycle_service import (
     PlanLifecycleService,
 )
 from cleveragents.config.settings import Settings
-from cleveragents.core.exceptions import NotFoundError, PlanError
+from cleveragents.core.exceptions import (
+    BusinessRuleViolation,
+    NotFoundError,
+    PlanError,
+    ResourceNotFoundError,
+)
 from cleveragents.domain.models.core.checkpoint import RollbackResult
 from cleveragents.domain.models.core.plan import (
     PlanPhase,
@@ -45,14 +50,19 @@ _NONEXISTENT_PLAN_ID = "01NONEXISTENT0000000000000"
 def _make_mock_checkpoint_service(
     plan_id: str = _PLAN_ID,
     checkpoint_id: str = _CHECKPOINT_ID,
+    *,
+    side_effect: Exception | None = None,
 ) -> MagicMock:
     """Build a mock CheckpointService that returns a valid RollbackResult."""
     mock_svc = MagicMock()
-    mock_svc.selective_rollback.return_value = RollbackResult(
-        restored_files_count=3,
-        changed_paths=["src/foo.py", "src/bar.py", "tests/test_foo.py"],
-        from_checkpoint_id=checkpoint_id,
-    )
+    if side_effect is not None:
+        mock_svc.selective_rollback.side_effect = side_effect
+    else:
+        mock_svc.selective_rollback.return_value = RollbackResult(
+            restored_files_count=3,
+            changed_paths=["src/foo.py", "src/bar.py", "tests/test_foo.py"],
+            from_checkpoint_id=checkpoint_id,
+        )
     return mock_svc
 
 
@@ -154,6 +164,38 @@ def step_plr_service_with_checkpoint_no_event_bus(context: Context) -> None:
     context.plr_checkpoint_id = _CHECKPOINT_ID
 
 
+@given(
+    "plr-a plan lifecycle service with a checkpoint service that raises ResourceNotFoundError"
+)
+def step_plr_service_checkpoint_raises_resource_not_found(context: Context) -> None:
+    """Set up a service whose checkpoint service raises ResourceNotFoundError."""
+    context.plr_mock_checkpoint_svc = _make_mock_checkpoint_service(
+        side_effect=ResourceNotFoundError(
+            "Checkpoint not found",
+            resource_type="checkpoint",
+            resource_id=_CHECKPOINT_ID,
+        )
+    )
+    context.plr_service = _create_service(
+        checkpoint_service=context.plr_mock_checkpoint_svc,
+    )
+    context.plr_checkpoint_id = _CHECKPOINT_ID
+
+
+@given(
+    "plr-a plan lifecycle service with a checkpoint service that raises BusinessRuleViolation"
+)
+def step_plr_service_checkpoint_raises_business_rule(context: Context) -> None:
+    """Set up a service whose checkpoint service raises BusinessRuleViolation."""
+    context.plr_mock_checkpoint_svc = _make_mock_checkpoint_service(
+        side_effect=BusinessRuleViolation("Checkpoint rollback failed"),
+    )
+    context.plr_service = _create_service(
+        checkpoint_service=context.plr_mock_checkpoint_svc,
+    )
+    context.plr_checkpoint_id = _CHECKPOINT_ID
+
+
 @given("plr-a plan in Execute phase with a sandbox")
 def step_plr_plan_in_execute_with_sandbox(context: Context) -> None:
     """Create a plan in Execute/PROCESSING state."""
@@ -334,6 +376,30 @@ def step_plr_not_found_error_raised(context: Context) -> None:
     )
 
 
+@then("plr-a ResourceNotFoundError should be raised")
+def step_plr_resource_not_found_error_raised(context: Context) -> None:
+    """Assert that a ResourceNotFoundError was raised."""
+    assert context.plr_raised_error is not None, (
+        "Expected a ResourceNotFoundError to be raised but no error was raised."
+    )
+    assert isinstance(context.plr_raised_error, ResourceNotFoundError), (
+        f"Expected ResourceNotFoundError, got {type(context.plr_raised_error)}: "
+        f"{context.plr_raised_error}"
+    )
+
+
+@then("plr-a BusinessRuleViolation should be raised")
+def step_plr_business_rule_violation_raised(context: Context) -> None:
+    """Assert that a BusinessRuleViolation was raised."""
+    assert context.plr_raised_error is not None, (
+        "Expected a BusinessRuleViolation to be raised but no error was raised."
+    )
+    assert isinstance(context.plr_raised_error, BusinessRuleViolation), (
+        f"Expected BusinessRuleViolation, got {type(context.plr_raised_error)}: "
+        f"{context.plr_raised_error}"
+    )
+
+
 @then("plr-a PLAN_ROLLED_BACK domain event should have been emitted")
 def step_plr_plan_rolled_back_event_emitted(context: Context) -> None:
     """Assert that a PLAN_ROLLED_BACK event was emitted."""

src/cleveragents/application/services/plan_lifecycle_service.py

@@ -18,7 +18,7 @@ Terminal outcomes live in Apply's processing state: ``applied``,
 | ``use_action(...)``       | Action (available)   | Strategize/QUEUED|
 | ``execute_plan(...)``     | Strategize/COMPLETE  | Execute/QUEUED   |
 | ``apply_plan(...)``       | Execute/COMPLETE     | Apply/QUEUED     |
-| ``rollback_plan(...)``    | Execute (any state)  | Execute/QUEUED   |
+| ``rollback_plan(...)``    | Any (non-terminal)   | *(unchanged)*     |
 
 ## Apply Terminal Outcomes
 
@@ -2149,7 +2149,10 @@ class PlanLifecycleService:
             current_state=plan.processing_state.value,
         )
 
-        # Delegate to CheckpointService for the actual sandbox rollback
+        # Delegate to CheckpointService for the actual sandbox rollback.
+        # Rollback restores sandbox state but does not mutate the plan's
+        # phase or processing_state; callers observe the pre-rollback
+        # lifecycle metadata.
         result = self.checkpoint_service.selective_rollback(plan_id, checkpoint_id)
 
         self._logger.info(