'Final answer' is often the final answer. 😅 - How to fix it?

[skye0402]

I have 2 or 3 agents chatting using gpt-4o. At then end a report is prepared as I requested it. However, the final answer in 50% of the cases looks like this:

<... lots of report text>
#### Conclusion
By leveraging the recommended services and .... bla bla....

#### References
1. [API_INSPECTIONPLAN_SRV](https://api.sap.com/api/API_INSPECTIONPLAN_SRV)
2. [API_CHARCATTRIBUTECATALOG_SRV](https://api.sap.com/api/API_CHARCATTRIBUTECATALOG_SRV)
3. [SAP Cloud SDK for S/4HANA](https://help.sap.com/doc/9b126cf98d314c02acc9fe9e3631b2da/1.0/en-US/com/sap/cloud/sdk/s4hana/datamodel/odata/services/package-summary.html)

Final Answer: the final answer to the original input question

> Finished chain.
 [2024-08-17 20:56:06][DEBUG]: == [Crew Manager] Task output: the final answer to the original input question

So the final answer is... drumroll...

the final answer to the original input question 😂

Needless to say I'm not interested in that. Anyone knows how to fix it? As said, it happens ~30-50% of all cases, not all the time. Crew is hierarchical.