Add endpoint monitor catalog and warn-only drift reconciler

Operators need to see when a machine's running services or processes fall outside the desired state defined by policy. This adds a reusable monitor catalog (shipped read-only rows plus user-built local rows), computes drift server-side on agent check-in, and surfaces a warning on the endpoint panels. Remediation stays manual for now; automation can follow once demand is proven. Includes migrations 0006-0008 and the monitors API client.

plugins/endpoints/commands.py

@@ -185,6 +185,36 @@ def requires_target(action):
     return action_targetkind(action) is not None
 
 
+# Monitor attachment as a policy action. A monitor definition is bound to a PC
+# type by carrying a ConfigurationPolicyAction whose name is 'monitor:<id>'.
+# Distinct names let one policy attach several monitors and fold naturally
+# (first-match-wins per name), and they are validated in the policy layer but
+# never enqueued as commands because ALLOWED_ACTIONS_FOR intersects with
+# COMMAND_ALLOWLIST, which carries no monitor: entries.
+MONITOR_ACTION_PREFIX = 'monitor:'
+
+
+def monitor_action_name(endpointmonitorid):
+    """Build the policy action name that attaches a monitor definition."""
+    return f'{MONITOR_ACTION_PREFIX}{endpointmonitorid}'
+
+
+def is_monitor_action(action):
+    """True when action names a monitor attachment (monitor:<id>)."""
+    return isinstance(action, str) and action.startswith(MONITOR_ACTION_PREFIX)
+
+
+def monitor_id_from_action(action):
+    """The monitor id an action attaches, or None when it is not a monitor action."""
+    if not is_monitor_action(action):
+        return None
+    raw = action[len(MONITOR_ACTION_PREFIX):]
+    try:
+        return int(raw)
+    except (TypeError, ValueError):
+        return None
+
+
 def ALLOWED_ACTIONS_FOR(pctype=None, assetid=None):
     """The action names an endpoint may run.
 
@@ -211,6 +241,44 @@ def ALLOWED_ACTIONS_FOR(pctype=None, assetid=None):
         # behavior). This is intentionally permissive; deny-by-default would be
         # a separate, explicit setting.
         return base
+    # Attaching a monitor expresses intent to watch a process/service AND to
+    # control it: the panel renders that row's Start/Stop/Status buttons, so the
+    # gate must permit the matching generic primitives or every click 409s
+    # "not permitted by policy". Expand each monitor:<id> to its kind's control
+    # actions before intersecting.
+    allowed = allowed | _monitor_control_actions(allowed)
     # Never widen past the code-level allowlist: policy can only gate within it.
     return base & allowed
 
+
+def _monitor_control_actions(action_names):
+    """Control primitives granted by the monitor: attachments in a policy.
+
+    A process monitor grants process_start/stop/status; a service monitor grants
+    service_start/stop/status. The monitor's kind is read from the catalog, so a
+    definition's kind is the single source of truth for what its row can do.
+    """
+    ids = set()
+    for name in action_names:
+        if isinstance(name, str) and name.startswith(MONITOR_ACTION_PREFIX):
+            monitor_id = monitor_id_from_action(name)
+            if monitor_id is not None:
+                ids.add(monitor_id)
+    if not ids:
+        return set()
+
+    try:
+        from .models import EndpointMonitor
+    except Exception:
+        return set()
+
+    granted = set()
+    for monitor in EndpointMonitor.query.filter(
+            EndpointMonitor.endpointmonitorid.in_(ids)).all():
+        kind = (monitor.kind or '').lower()
+        if kind == 'process':
+            granted |= {'process_start', 'process_stop', 'process_status'}
+        elif kind == 'service':
+            granted |= {'service_start', 'service_stop', 'service_status'}
+    return granted
+