Detecting and Containing Runaway Agent Loops
Frontier models burn through tokens faster in loops, and most teams can't stop them once they start.
Diego Lindqvist
Section
Features
3 stories in Features.
Prompt Injection Attack Surfaces in Production Agents
Autonomous agents reading untrusted content face a production-scale prompt injection problem.
Rohan Sørensen
Circuit Breakers for AI Agent Tool Calls
Automated safeguards stop runaway agents before they drain accounts or damage reputation.
Mateo Halvorsen