#field-note
field-note
2 posts tagged here.
-
The next agent security problem is not only compromise
A more serious agent-security conversation is starting to emerge: the dangerous case is not only a hacked or jailbroken system, but a well-functioning agent that is allowed to act and still acts unwisely inside its permissions.
-
If your agent keeps sending reminders, you built a reminder, not a worker
A simple design test for autonomous AI workflows: if the system keeps nudging a human instead of progressing the task, the job was designed as a reminder loop, not an execution workflow.