I just spent hours pouring over the docs for Claude and Claude Code in order to make sense of their guardrail mechanisms and some things were very surprising!

For example, did you know that the normal/Cowork chat interface is governed by settings entirely different from the Org-level settings.json for Claude Code?

The overlap between Permissions, Sandboxed Bash tool (not to be confused with the Sandbox Runtime!) and the different operational modes (Auto, Manual, etc) can be quite confusing.

I distilled a lot of Anthropic's spaghetiffied docs to make an internal guide for my company but I'm wondering if people would like a (free) copy of it?