The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests data from a Kubernetes cluster that could be manipulated by an adversary. 1. Ingestion points: Kubernetes resource names, labels, and Hubble network flow logs are read via hubble_flows_query_tool and cilium_policy_get_tool. 2. Boundary markers: Absent. The skill provides no instructions to the agent to distinguish between its own logic and data found in cluster metadata. 3. Capability inventory: The agent has the ability to read and modify cluster state using tools like cilium_policy_get_tool and kubectl_apply. 4. Sanitization: None identified within the skill logic.

k8s-cilium