litmus-chaos

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill contains runtime install commands that fetch and apply remote executable content (Helm chart and Kubernetes manifest) from https://litmuschaos.github.io/litmus-helm/ and https://hub.litmuschaos.io/api/chaos/2.14.0?file=charts/generic/pod-delete/experiment.yaml, which will download and deploy code into the cluster at runtime.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill instructs deploying and running destructive Kubernetes chaos experiments (e.g., kubelet-service-kill, node-drain, disk-fill, pod-delete) and creating/assigning Kubernetes service accounts/RBAC, which can modify cluster and node state and disrupt machines the agent controls.