Thanks Pawel! Yes I noticed you had tweaked it since I started writing this, and only gave your new prompt a mention at the end. Should I replace your first prompt with the one at this link? https://substack.com/@huryn/note/c-228204100?r=1to7jv
Excellent post! Could you clarify whether you put Pawel's prompt in ~/.claude/CLAUDE.md, or in CLAUDE.md at the project level? I assume that the file structure you discuss is global, but that is not always 100% clear. Here is an example: in CLAUDE.md, you added:
```
Maintain a structured memory system rooted at .claude/memory/
```
I assume you mean ~/.claude/memory (notice the tilde + slash.) Just want to make sure. As models become better at following instructions, lack of precision will lead to errors in the future where none existed before. Thanks.
Hey Gordon! I put it in ~/.claude/CLAUDE.md correct
For the excerpt you have there - I cannot for the life of me remember if that is intentional and I was trying to make it work for both a user creating a project Claude.md or a global one or if this is a mistake 😬
Now I am worried if I take it out it will have side effects 😀 It has persisted into my own memory even after doing a similar prompt in plan mode so Claude must think it makes sense
> lets create a brand new skill. Lets call it /memory-manage and it will take initialise, organise, optimise Paste: Above prompt, Paste Above Hook. I dont want to actualy create a hook I would rather run that capability manually using a comand like /memory-manage organise
Stellar job, John. And I should have more carefully verified the prompt. My actual solution uses... index 😀https://www.productcompass.pm/p/self-improving-claude-system
Thanks Pawel! Yes I noticed you had tweaked it since I started writing this, and only gave your new prompt a mention at the end. Should I replace your first prompt with the one at this link? https://substack.com/@huryn/note/c-228204100?r=1to7jv
Organizing Claude memory with clear categories is a game changer. Well documented.
Hi John,
Excellent post! Could you clarify whether you put Pawel's prompt in ~/.claude/CLAUDE.md, or in CLAUDE.md at the project level? I assume that the file structure you discuss is global, but that is not always 100% clear. Here is an example: in CLAUDE.md, you added:
```
Maintain a structured memory system rooted at .claude/memory/
```
I assume you mean ~/.claude/memory (notice the tilde + slash.) Just want to make sure. As models become better at following instructions, lack of precision will lead to errors in the future where none existed before. Thanks.
Hey Gordon! I put it in ~/.claude/CLAUDE.md correct
For the excerpt you have there - I cannot for the life of me remember if that is intentional and I was trying to make it work for both a user creating a project Claude.md or a global one or if this is a mistake 😬
Now I am worried if I take it out it will have side effects 😀 It has persisted into my own memory even after doing a similar prompt in plan mode so Claude must think it makes sense
> lets create a brand new skill. Lets call it /memory-manage and it will take initialise, organise, optimise Paste: Above prompt, Paste Above Hook. I dont want to actualy create a hook I would rather run that capability manually using a comand like /memory-manage organise
It's on the agenda Baz! I'm getting married next month so my free time is non-existent at the moment 😀