k8s-service-mesh
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill possesses write/execute capabilities through
kubectl_applyandapply_manifesttools. - Ingestion points: The skill instructions demonstrate taking YAML content (e.g., VirtualService, DestinationRule) and applying it to the cluster (SKILL.md, TRAFFIC-SHIFTING.md).
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the manifests are present in the skill files.
- Capability inventory:
kubectl_apply,apply_manifest, andistioctl installprovide the agent with the ability to modify cluster state and network routing. - Sanitization: There is no evidence of sanitization or validation logic for the YAML manifests before they are applied.
- Command Execution (MEDIUM): The skill documentation includes shell commands for installation (
istioctl install --set profile=demo). While documented as a prerequisite, an agent following these instructions might execute commands in a privileged context without sufficient validation of parameters.
Recommendations
- AI detected serious security threats
Audit Metadata