[Bug] Anthropic API Error: Unexpected reasoning_extraction block rejection on valid prompt

Bug Description
I ran the following prompt to re-verify the review bot's comments:

## Context:

Are you sure your findings and the solution are all correct? You need to give me stronger evidence. Don't be biased toward either the reviewer's side or the opposite side; be objective.


## Tasks:

1. Critically re-examine each (findings, the solution) and give me the results with very strong evidence.

- Make sure to verify your findings by running the existing tests, creating temporary test files, and/or amending the code. Once the verification is complete, tell me what you did in the final response, and remove any temporary files and/or revert the code to its original state.



2. Based on the results of Task 1, if each is correct or partially correct so that I need to make changes in the current code, give me the best solution.

- Make sure to verify your findings by running the existing tests, creating temporary test files, and/or amending the code. Once the verification is complete, tell me what you did in the final response, and remove any temporary files and/or revert the code to its original state.

- The solution must include before/after code, file paths, and line numbers. 



## Requirements:

- IF there is something to be clarified, ask/grill the user.

- Always be objective, don't be biased to any side (e.g. previous PR review comment, your previous suggestion, my previous argument... etc.)

- For each task, make sure to verify your findings by running the existing tests, creating temporary test files, and/or amending the code. Once the verification is complete, tell me what you did in the final response, and remove any temporary files and/or revert the code to its original state.

- The solution must include before/after code, file paths, and line numbers.

- Check thoroughly and give the best, simplest, clearest and safest answer to each question. Don't overkill.

- Always give me clear evidence with file name and code.

- Scan ALL the related files NEWLY so that the analysis is based on the latest codebase

**- First,
list the reasoning you considered in order to form your answer**, then evaluate for yourself whether that reasoning is reliable. After that, **summarize only the final conclusion separately at the end.**
If the evidence is weak, clearly state:
"This may not be accurate."

**- Whenever possible, explicitly mention the basis for your statements, using expressions such as: "Based on —" or "It is generally known that -". When referring to specific years, figures, names, or places, also indicate: **Accuracy: High / Medium / Low**

- Give me the good starting points about where to look up to understand the tasks in codebase and web. Make sure to verify the reliability of the information sources, and check if each hyperlink is accurate and accessible.



## Constraints/Restrains: 

- Do NOT modify any code files directly, provide the solution as guidance only (except creating the files to test your solution for getting empirical evidence).

- **Do NOT make up information you do not know under any circumstances.** If the information is beyond the provided materials or outside the scope of your knowledge, say:
"This information goes beyond the provided materials or my knowledge scope."
If speculation is required, make sure to state clearly at the beginning:
"This is speculation."


## Final instructions: 
- Make sure all the things (e.g. the role, the tasks, the context,
the restraints, the requirements, etc.) are all thoroughly fulfilled before finalising the response

I have been using this prompt with no issue until a few hurs ago. However, it suddenly gets blocked:

 API Error: Sonnet 5 can't help with this. Start a new session to continue.

  Send feedback with /feedback or learn more: https://www.anthropic.com/legal/aup

  Details: `[reasoning_extraction]`

  Request ID: req_011CfNdAHAF7pArVG4GGwA9J

I am not sure why my prompt is considered as "reasoning extraction", it has no such thing.

Environment Info

• Platform: darwin
• Terminal: vscode
• Version: 2.1.281
• Feedback ID: cefd67e2-76b0-4b99-8c5a-43f4903aac70

Errors

[]