AGENT_CASCADING_FAILURES
Verify AI agents have safeguards against cascading failure scenarios: retry limits, execution sandboxing, error handling, and resource constraints.
Framework Mappings
OWASP Agentic ASI08 — Cascading FailuresEU AI Act Article 9 — Risk Management SystemNIST AI RMF MP5 — Impacts Assessed
Agent Frameworks Scanned
Claude CodeLangChainCrewAIAutoGenMCP
Evaluation Facets
01
Retry & Loop Limits
RequiredWeight: 0.35
Pass:Configured retry limits or circuit breakers
Fail:No loop limits — can retry indefinitely
02
Execution Sandboxing
RequiredWeight: 0.3
Pass:Code execution sandboxed (Docker, scoped Bash, deny rules)
Fail:Unsandboxed execution
03
Error Handling & Degradation
Weight: 0.2
Pass:Graceful error handling with deny rules and human checkpoints
Fail:No error handling
04
Resource Limits
Weight: 0.15
Pass:Resource constraints on agent operations
Fail:Unconstrained operations
Remediation Steps
- 1Configure retry limits and circuit breakers to prevent infinite loops
- 2Sandbox execution to contain failures
- 3Add graceful error handling with degradation paths
- 4Set resource constraints on all agent operations
Evaluate this control automatically
Connect your repos and Brigs evaluates AGENT_CASCADING_FAILURES across all your agent configurations.
Get Started — Free