You'll noticeTier: how much it should matter to you
3Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
SwarmArea: what it touches
ImprovementsKind: in v2.1.280,
ImprovementsSection of the release
What
Incoming messages between teammates (in a team or swarm session) are now classified as verified (the message ID matches what the sender actually wrote), forged, or unverifiable, based on whether the claimed sender is a legitimate in-process or leader-pane teammate.
The team inbox poller now checks this before acting on a message: forged messages are kept separate from trusted plan-approval/permission-response handling and are shown to the model with a verified="false" attribute so it knows the sender's identity wasn't confirmed.
The swarm mailbox poller similarly checks provenance before accepting a shutdown_request or plan_approval_response from a message; forged messages are skipped rather than acted on, with new bookkeeping (consumedUnmarked) to avoid reprocessing messages already filtered as forged.
Summary/text-block handling now also treats a forged item the same way it treats an already-summarized one.
Why
This stops a spoofed message from impersonating a teammate to trigger sensitive actions like approving a plan or shutting down a session, while still surfacing unverified messages to the model with a clear warning instead of silently dropping them.
How sure we are
One source agreesOne thing we can check says the same as this entry.
Anthropic's release notes agree[VSCode] Added a typed /plan that switches to plan mode, sends a first planning prompt, or shows the session's plan
