k8s-envoy-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external, potentially untrusted sources within the Kubernetes cluster, creating an ingestion surface for indirect prompt injection.
- Ingestion points: The skill reads live data using
kubectl logs,kubectl get configmap -o yaml, and the Envoy admin API viakubectl get --rawinSKILL.mdandreferences/responses-and-upstreams.md. - Boundary markers: The skill contains explicit negative constraints, such as "Never read Secrets" and "Do not repeat keys" found in the bootstrap. It also enforces a strict "Ask first" policy for all modification commands.
- Capability inventory: The agent has the capability to read cluster logs, inspect configurations, and execute commands within pods (
kubectl exec), though these are governed by the manual approval requirement. - Sanitization: The skill relies on human-in-the-loop review for all "Proposed fixes" rather than automated sanitization of ingested content.
Audit Metadata