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
Product
TS-Bench
- Mentions · 7d0
- Mentions · 30d1
- All mentions1
- First seenOct 16, 2026
- Last mentioned9d ago
Timeline
1 articles · newest firstAppears with
- TS-Guardproduct · 1
- StepGuardproduct · 1
- ToolRosettaproduct · 1
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.