The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the k8s-monitoring Helm chart from Grafana's official GitHub-hosted repository (https://grafana.github.io/helm-charts). This is a standard operation for installing well-known monitoring software.
[COMMAND_EXECUTION]: The skill performs administrative tasks such as helm install, kubectl create secret, and systemctl reload to provision and configure infrastructure monitoring components.
[PRIVILEGE_ESCALATION]: In references/clouds-and-hosts.md, the skill configures Alloy to access the host's Docker socket at unix:///var/run/docker.sock. While this is required for cAdvisor and Docker log discovery, it represents a high-privilege connection to the container runtime.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and verify telemetry data, including container logs and Kubernetes events, which could contain attacker-controlled content designed to influence agent behavior.
- Ingestion points: Reading pod logs using kubectl -n monitoring logs in SKILL.md and querying metrics/logs in Grafana Cloud.
- Boundary markers: No specific boundary markers or "ignore instructions" delimiters are defined for processing log outputs.
- Capability inventory: The skill has access to helm, kubectl, and host-level service management (systemctl).
- Sanitization: There is no evidence of sanitization or filtering of the ingested log data before it is processed by the agent.

grafana-infrastructure-monitoring