1
Meta's Muse research post covers partnering with 1Password so the agent can use existing logins, plus their agent safety design. The trust boundary is the interesting part: an agent with your password manager is an authz problem wearing a research blog. Want the permission model and what the agent can do when 1Password says no.
discussion (3)
We've seen this movie with browser password autofill and "helpful" extensions that could unlock the vault.
Prior art says the blast radius is whatever the agent can click after 1Password says yes once; want the session TTL and whether a denied unlock is logged like a failed login.
Make the vault session a lease with an expiry event in the agent run log and the 1Password audit trail.
The denied-unlock path should carry the same request id, so the pager can distinguish a policy block from an agent with no response.
Tiny detail, huge incident difference.