The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and fetches Helm charts and Docker images from official repositories maintained by Grafana and the Prometheus community.
[DATA_EXFILTRATION]: Configurations include mounting system paths such as /var/run/docker.sock, /proc, and /sys to enable service discovery and metric collection. Additionally, the host root filesystem is mounted as read-only for hardware metric gathering. These mounts are standard operational requirements for monitoring agents and exporters.
[CREDENTIALS_UNSAFE]: Example configurations use descriptive placeholders (e.g., YOUR_ID/YOUR_TOKEN, your-app-password) for sensitive fields. The documentation provides clear instructions for users to manage their actual credentials securely using .env files and native Kubernetes Secrets.
[PROMPT_INJECTION]: The skill describes a monitoring system that ingests log and metric data from containers and the host, representing an indirect prompt injection surface.
Ingestion points: System logs and metrics are aggregated from the entire infrastructure into the Loki and Prometheus backends (SKILL.md, references/compose-deploy.md).
Boundary markers: No specific delimiters or isolation markers for ingested log content are defined in the provided configuration templates.
Capability inventory: The skill defines monitoring infrastructure and does not grant the agent additional execution or system-level capabilities.
Sanitization: External log and metric data are ingested directly into the monitoring stack without specified filtering or sanitization steps in the reference configurations.

homelab-monitoring