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-Guard
- Mentions · 7d0
- Mentions · 30d2
- All mentions2
- First seenOct 15, 2026
- Last mentioned9d ago
Timeline
2 articles · newest firstShortcut Sensitivity in Agent-Safety Evaluations
Shortcut cues in agent-safety benchmarks can change guard decisions and measured accuracy.
48
Appears with
- StepGuardproduct · 1
- AgentHarmproduct · 1
- TS-Benchproduct · 1
- LinuxArenaproduct · 1
- ToolRosettaproduct · 1
- Haikuai model · 1
- AgentDojoproduct · 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.