loki-logging
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The provided docker-compose.yml example sets GF_AUTH_ANONYMOUS_ENABLED=true and GF_AUTH_ANONYMOUS_ORG_ROLE=Admin. This configuration is insecure as it grants full administrative privileges to any user with network access to the Grafana instance without requiring a login.
- [DATA_EXPOSURE]: The skill configuration mounts sensitive host paths into containers. It accesses the host Docker socket (/var/run/docker.sock) and mounts system log paths (/var/log and /var/lib/docker/containers). While standard for log aggregation, these provide a broad surface for accessing sensitive system information.
- [EXTERNAL_DOWNLOADS]: The skill downloads components from official and well-known sources. It adds the official Grafana Helm repository at https://grafana.github.io/helm-charts and uses official Docker images for Loki, Promtail, and Grafana.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion surface for untrusted external data that could contain malicious instructions. 1. Ingestion points: Scrapes system logs, application logs, and Docker container logs. 2. Boundary markers: The LogQL and Promtail configurations do not include delimiters or instructions to ignore embedded commands in the log stream. 3. Capability inventory: The skill enables log querying via LogQL and deployment via Helm and Docker Compose. 4. Sanitization: No sanitization or filtering of log content is implemented in the provided pipeline stages.
Audit Metadata