grafana-observability
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches and runs the Grafana MCP server from the official Grafana GitHub repository using the 'uvx' tool.
- [COMMAND_EXECUTION]: Executes the 'uvx' command to launch the observability MCP server in the local environment.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including Loki logs, Prometheus metrics, and dashboard definitions which could potentially contain malicious instructions.
- Ingestion points: 'query_loki_logs', 'query_prometheus', 'get_annotations', and 'get_dashboard_by_uid' in 'SKILL.md'.
- Boundary markers: The skill includes specific rules to avoid secrets in queries but lacks explicit prompt delimiters or instruction-bypass warnings for external data.
- Capability inventory: Write-access tools include 'update_dashboard', 'patch_dashboard', 'create_alert_rule', 'update_alert_rule', 'create_incident', and 'add_activity_to_incident' as defined in 'SKILL.md'.
- Sanitization: No explicit content sanitization, filtering, or validation logic is specified for the ingested external telemetry data.
Audit Metadata