A symlink is a file that points to another file or folder. Claude Code follows these links to find the real location. The code that does this now handles two cases differently:
A chain of links that goes past 64 hops, which usually means a loop.
A link whose target sits on a network location.
In these cases it now returns the original path or nothing, rather than passing the raw result through. When the resolver has been asked to surface targets that go up a folder with .., it now raises an error instead.
Why
Links that loop forever or point off to network locations no longer slip through as if they were ordinary resolved paths.
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtThe finding does not say which cases return the original path and which return nothing.
Anthropic's release notes agreeImproved screen-reader output in /mcp: a disabled server is read as "off" instead of "pending"
