Tool-call guards fail when actions are rendered as shell commands
Four of five specialized guards raise AgentDojo false-positive rates by 21 to 47 points under shell form.
58
Four of five specialized guards raise AgentDojo false-positive rates by 21 to 47 points under shell form.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 2.2K readers · 2.7K views · 400 reads · all time 108K views