Fix 09c_hitl_streaming: approval pause never fires

The agent's instructions said "Only suggest deleting data if explicitly
asked", so the model returned the approval question as its final answer
(finishReason: STOP) instead of calling the tool. No WAITING event was
emitted, so the example's approval handler never ran and the
approval_required=True pause it exists to demonstrate never fired.

Reworded as ordered steps that name the tool and leave approval to the gate.
Verified 4/4 on openai/gpt-4o-mini and anthropic/claude-sonnet-4-5:
check -> restart -> approval pause -> delete.

A one-word suggest->delete change is not enough: any prose phrasing states
the deletion as a permission, and gpt-4o-mini treats a permission as
optional — it restarts, reports success, and skips the tool (0/5 across two
prose variants). Numbered imperative steps get it executed.

Author: ling-senpeng13

examples/agents/09c_hitl_streaming.py

@@ -41,9 +41,14 @@ def delete_service_data(service_name: str, data_type: str) -> dict:
     model=settings.llm_model,
     tools=[check_service, restart_service, delete_service_data],
     instructions=(
-        "You are an operations assistant. You can check, restart, and manage services. "
-        "If a service is unhealthy, check it first, then restart it. Only suggest "
-        "deleting data if explicitly asked."
+        "You are an operations assistant. Work through the request one tool call at a "
+        "time, in this order:\n"
+        "1. Check the service with check_service.\n"
+        "2. If it is unhealthy, restart it with restart_service.\n"
+        "3. Last, if the user asked you to clear or delete data, call "
+        "delete_service_data.\n"
+        "A human approves the deletion, not you — delete_service_data pauses for that "
+        "approval by itself, so never ask for approval in your own reply."
     ),
 )