The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides structured kubectl commands to inspect Kubernetes resources, including pods, events, and application logs. These commands are essential for platform diagnostics.
[DATA_EXFILTRATION]: Instructions are included to inspect environment variables (e.g., DB_HOST, AZURE_APP_CLIENT_ID) and decode authentication tokens. While this provides access to configuration details, the operations are intended for local debugging within the user's authorized environment and do not involve exfiltration to external third-party domains.
[EXTERNAL_DOWNLOADS]: Includes commands that use wget to verify connectivity to official Azure Active Directory JWKS endpoints. These are well-known services used for identity management.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it processes application logs and Kubernetes events. Ingestion points: kubectl logs and kubectl get events. Boundary markers: None. Capability inventory: Execute arbitrary shell commands via kubectl and shell pipes. Sanitization: None. This is a common characteristic of troubleshooting tools that analyze external system output.

nav-troubleshoot