# Starting Today, my hook forcing Claude to understand me and what i wanted started getting blocked by [reasoning extraction claude]. Does someone have a methold to solve the issue that my hook was previously solving ? — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/f698e7d34ca87616

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-11-03 10:37 ingested 2026-11-04 02:37

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xmpmg8/starting_today_my_hook_forcing_claude_to/)

## retained text



```
Until Yesterday, I had a prompt hooked when i was writting specific trigger. 

But since Today, it's started get blocked everytimes. I tried reword differently or in analysis, result. But doing so was loosing all the efficiency of this hook.

I understand why it's there. But my problem is that without it, i end up coming back to same shortcoming of Claude everytime : Not re-reading the conversation, focus on quick problem solving and quickly forgetting or skipping my previous demand.   
  
Especially when sometimes, the context just accumulate, even if i use sub agent, even if i use document as intermediary to save the progress.



**## The Prompt**

>\##Pre-work

>Before Answering, Start a reflection on what I said and want, re-reading the conversation

>\- 6 Paragraphs Minimum

>\- The Reflection must show the process of understanding ME trough this session

>\- The Reflections must show the process of understanding WHAT I WANT





**## The goal**

I was using this for multiple reasons :

1) I was forcing a re-read of the previous message, instead of focusing mostly on the lasts message when the context of the conversation was growing and it was more important. 

2) I was focusing, not a solving the issue, but understanding what i was asking. Because a lot of the time, it did assumption about what i wanted, drifted and slight changed the goal or used his own tough about what has to be done. While with this method, it always understood what i was asking.

3) I wanted to avoid empty answer that did'nt amount to anything.  
I'm not interested in Claude ackwnoledge mistake, this was just empty word who drove no change, i wanted him to understand and continue from what i wanted.

4) Finally, I was asking for the reflexion because i did'nt want him to be in "plan mode" or "problem solving". I wanted him to be in "human understanding"  
  
I often have the problem after a request, he jump on solving the problem immediately and he does it the wrong way or by ignoring prior correction. It's ended up solving an issue, redoing one prior, etc.



This is where the reflection was helpfull, because he was building on it and he demonstrated understanding of what i want. I was using the fact that he wrote it so it was the last token present



**## Question**

So I was using this in order to overcome the shortcoming and being able to resume my work on good base.

But now, i'm blocked everytime i use the hook. And every alternative end up with him focus on solving problem, not reading the past. I even tried to make a sub agent analysis the problem, but he dropped what was told quickly.

Does someone has a good way to solve thoses issues ?
```



document fetched 2026-11-04 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xmpmg8/starting_today_my_hook_forcing_claude_to
