DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.
Auto-mode decisions are cached by how long they stay valid
Under the hood
Useful2Signal3
Permissions
Cached auto-mode verdicts now expire by how long they stay valid, and long-transcript outcomes get reused.
What's wrong with this entry?
Anonymous. No account, no email.
What
The cached per-tool verdict now records a reuse policy instead of just whether it was a denial. Denials are reused indefinitely; approvals, and now also "transcript too long" outcomes that used to be thrown away, are reused only while the conversation has not moved on, tracked by message count and the last message's id.
Evidence
same-transcript
Strings lifted out of the shipped bundle, so the claim above can be checked against them.