Invalid tool call parsing in thinking token for hermes format in tool_parser.py in agent_loop from experiment folder.

[chetandhembre]

Hello
in HermesToolParser class.. extract_tool_calls function does eliminate content from <think>..</think> tags for searching for tool call signature..
this lead to case where tool call syntax used in thinking tokens consider as valid tool call.
example (Formatting for readability purposes)

<think> 
this is just my reasoning.. let me call tool A 
<tool_call>{"name": "A", "arguments": {}}</tool_call> 
it will give me result which i will in my analysis
</think>
<tool_call>{"name": "A", "arguments": {}}</tool_call>

in above dummy example tool_call is used in side thinking token and out side also.. parser will consider this as two tool call..
I am wondering is it correct behaviour?