Summary
While reproducing the credential-store-split issue (see the linked issue below), both
bt trace doctor codex and bt trace doctor claude reported:
Auth: ready (saved_profile)
at the exact same time the daemon's log showed every incoming event being rejected due to a
credential mismatch. doctor validates the credential store it can see directly (the real one), not
the one the daemon actually resolves against (which can be the virtualized one under app
sandboxing) — so it reports a healthy configuration during total data loss.
Since doctor is the tool users are pointed to when tracing looks broken, this false positive
actively misleads troubleshooting.
Suggested fix
Make doctor resolve credentials through the same code path the daemon itself uses, so a mismatch
between what the CLI sees and what the daemon sees is detected rather than masked.
Related
#381 (the underlying credential-store split)
Summary
While reproducing the credential-store-split issue (see the linked issue below), both
bt trace doctor codexandbt trace doctor claudereported:at the exact same time the daemon's log showed every incoming event being rejected due to a
credential mismatch.
doctorvalidates the credential store it can see directly (the real one), notthe one the daemon actually resolves against (which can be the virtualized one under app
sandboxing) — so it reports a healthy configuration during total data loss.
Since
doctoris the tool users are pointed to when tracing looks broken, this false positiveactively misleads troubleshooting.
Suggested fix
Make
doctorresolve credentials through the same code path the daemon itself uses, so a mismatchbetween what the CLI sees and what the daemon sees is detected rather than masked.
Related
#381 (the underlying credential-store split)