Y
Hacker News
new
|
ask
|
show
|
jobs
by
xms17189
43 days ago
Curious how you handle trust boundaries for tool outputs here. Do you keep a signed or replayable trace so a developer can audit what the agent saw before it acted?