k8s-service-mesh

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from the Kubernetes cluster and has the capability to write changes back to the cluster. An attacker with the ability to modify Kubernetes resources could embed malicious instructions in metadata or configuration fields that the agent might follow when inspecting those resources.
  • Ingestion points: The skill reads cluster state using tools like istio_virtualservice_get, istio_virtualservices_list, istio_destinationrules_list, and istio_peerauthentications_list (found in SKILL.md, MTLS.md, and TRAFFIC-SHIFTING.md).
  • Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when processing the data from these tools.
  • Capability inventory: The skill includes the kubectl_apply tool which allows the agent to modify the cluster configuration (found in SKILL.md and TRAFFIC-SHIFTING.md).
  • Sanitization: No explicit sanitization, validation, or filtering of the ingested resource content is implemented before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 06:52 PM