Greeting discloses AVA is automated (HIPAA item); never claim to be human

CLAUDE.md compliance section requires AVA to identify as automated at call
start. Greeting now says "this is AVA, an automated assistant", and a prompt
guardrail makes her answer honestly if a caller asks whether she's an AI.
Replay harness greeting kept in sync.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

bot.py

@@ -202,6 +202,10 @@ SYSTEM_PROMPT = (
     "example, after noting insurance, in the SAME turn go straight on to ask the preferred day "
     "and time. A dead-end statement leaves the caller unsure whose turn it is and causes silence.\n\n"
     "Stay truthful and within your limits:\n"
+    "- You are an automated assistant and you say so in the greeting. If a caller asks whether "
+    "they're talking to a real person or an AI, answer honestly and briefly ('I'm an automated "
+    "assistant, but I can take your request and staff will follow up'), then carry on — never "
+    "claim to be human.\n"
     "- Use ONLY the facts below for addresses, phone numbers, insurance, and services. Never "
     "make any of these up.\n"
     "- OFFICE SELECTION: ask what city or area is most convenient. When the caller names a place "
@@ -752,10 +756,11 @@ async def run_agent(transport, caller_number=None, call_sid=None, do_capture=Tru
         logger.info("Client connected -- greeting")
         if audiobuffer:
             await audiobuffer.start_recording()
+        # HIPAA/compliance: AVA identifies as automated at call start — no exceptions.
         await task.queue_frames(
             [TTSSpeakFrame(
-                f"Thank you for calling Advanced Vision Care, this is {AGENT_NAME}. "
-                "How can I help you today?"
+                f"Thank you for calling Advanced Vision Care, this is {AGENT_NAME}, "
+                "an automated assistant. How can I help you today?"
             )]
         )