Claude Code: Fix error field handling - explicitly set null on healthy state

- app.py: Explicitly set error=None when imports succeed
- brain_minimal.py: Add note that error field is deliberately reset on state updates
- Both files: Add _note explaining error=null means healthy

This ensures the error field is always null (not "unknown" string) when Cain is healthy.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app.py

@@ -432,9 +432,14 @@ try:
         status_data["startup_checks"]["openclaw_imported"] = True
         from agents import brain_minimal
         status_data["startup_checks"]["brain_imported"] = True
+        # CRITICAL: Explicitly set error to null when imports succeed
+        status_data["error"] = None
     except Exception as e:
         status_data["error"] = f"{type(e).__name__}: {e}"
 
+    # Add explanatory note about error field semantics
+    status_data["_note"] = "error=null means healthy - null or 'unknown' string are both treated as 'no error'"
+
     # Write to primary location (OPENCLAW_DATA_DIR)
     status_path = os.environ.get('CAIN_STATUS_PATH', '/data/cain_status.json')
     Path(status_path).parent.mkdir(parents=True, exist_ok=True)

openclaw/.openclaw/agents/brain_minimal.py

@@ -240,10 +240,12 @@ class AgentStateMachine:
             "current_state": self._state.value,
             "last_updated": datetime.utcnow().isoformat() + "+00:00",
             "agent": self._agent_name,
-            "error": None  # Explicitly clear error when state machine updates
+            "error": None,  # CRITICAL: Explicitly clear error when state machine updates (null = healthy)
+            "_note": "error=null means healthy - treated as 'no error'"
         }
 
         # Preserve stage, a2a, and startup_checks fields if they existed
+        # NOTE: We deliberately DO NOT preserve the error field - always reset to None on state updates
         if existing_stage:
             status_data["stage"] = existing_stage
         if existing_a2a: