4 Comments
User's avatar
Fabrice Talbot's avatar

Such an interesting find!

I wish Anthropic would do a better job at documenting their product. Or pay us to do so 😂

I use Claude projects. My setup is as follows to minimize session start risks:

- bootstrap in project instructions

- read from project knowledge first then the instruction (claude.md or other) and fallback to claude.md (mine is in the cloud)

- bootstrap validates each instruction section was captured (~10, on exact string match so no approximation)

- once read, confirmation message that includes the EOF marker in claude.md (last line)

It doesn’t address the 10k but it provides a welcome read receipt to a business user.

Eugeniu Ghelbur's avatar

Thanks Fabrice. The read receipt with the last line of claude.md is a smart check, and exact string matching is the right way to do it. Have you seen a session where it failed to read everything?

Fabrice Talbot's avatar

Not yet because I added it this week and I am compliant so far but I’ve seen one where a typo on the claude.md session titles in the bootstrap would raise a failure (I list the ## titles to make sure they get matched too, I have <10)

Eugeniu Ghelbur's avatar

Fabrice, 1 typo tripped your check, so the exact match works. It catches a renamed section before Claude quietly skips it.

When it fails, does the session stop, or only warn you?