The Agent Skills Directory

[DATA_EXFILTRATION]: The Prometheus configuration template explicitly references sensitive Kubernetes service account files used for API authentication.
Evidence: Path references to /var/run/secrets/kubernetes.io/serviceaccount/token and /var/run/secrets/kubernetes.io/serviceaccount/ca.crt in references/prometheus-configuration.md.
[CREDENTIALS_UNSAFE]: The Alertmanager configuration contains a hardcoded password field with a default value.
Evidence: auth_password: "secret" in references/alertmanager-configuration.md.
[INDIRECT_PROMPT_INJECTION]: The health-check.sh script accepts a service URL as a command-line argument without performing any input validation or sanitization, creating a surface for potential injection if the script is expanded or executed with untrusted input.
Ingestion points: SERVICE_URL variable in scripts/health-check.sh sourced from the first command-line argument.
Boundary markers: Absent; the input is interpolated directly into shell commands.
Capability inventory: The script performs shell operations and is intended to interface with external network services.
Sanitization: No validation of the URL format or character escaping is performed.
[COMMAND_EXECUTION]: The deployment reference script executes multiple kubectl commands, including background port-forwarding which modifies the local network configuration.
Evidence: kubectl port-forward commands in references/monitoring-deployment.md.

infrastructure-monitoring