fix(claude-agent): warn on empty result, debug-log hook skips

- activities.py: log WARNING when Claude returns 0 messages (catches
  silent failures from model connectivity issues e.g. VPN/proxy down)
- hooks.py: log DEBUG when pre/post_tool_use hooks bail early due to
  missing task_id, making it visible why tool call logs aren't appearing

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

Author: gabrycina

src/agentex/lib/core/temporal/plugins/claude_agents/activities.py

@@ -145,6 +145,11 @@ async def run_claude_agent_activity(
         await handler.cleanup()
 
         results = handler.get_results()
+        if not results.get("messages"):
+            logger.warning(
+                f"[run_claude_agent_activity] Claude returned 0 messages — "
+                f"possible model connectivity issue (task_id={task_id}, workspace={workspace_path})"
+            )
         logger.debug(f"Returning results with keys: {results.keys()}")
         return results
 

src/agentex/lib/core/temporal/plugins/claude_agents/hooks/hooks.py

@@ -66,6 +66,7 @@ async def pre_tool_use(
             Empty dict (allow execution to proceed)
         """
         if not self.task_id or not tool_use_id:
+            logger.debug(f"Hooks skipping pre_tool_use — task_id={self.task_id}, tool_use_id={tool_use_id}")
             return {}
 
         tool_name = input_data.get("tool_name", "unknown")
@@ -133,6 +134,7 @@ async def post_tool_use(
             Empty dict
         """
         if not self.task_id or not tool_use_id:
+            logger.debug(f"Hooks skipping post_tool_use — task_id={self.task_id}, tool_use_id={tool_use_id}")
             return {}
 
         tool_name = input_data.get("tool_name", "unknown")