grafana
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes structured Grafana configuration data (dashboards, data sources, and alert rules) that may originate from untrusted sources, presenting an opportunity for instructions embedded in data to influence the agent.
- Ingestion points: The
Tools/DashboardCrud.tsandTools/GrafanaClient.tsfiles contain logic to read JSON files from the local file system and retrieve configurations directly from the Grafana HTTP API. - Boundary markers: The instructions do not implement explicit boundary markers or delimiters when processing JSON fields like dashboard descriptions, panel titles, or alert annotations, which could contain adversarial content.
- Capability inventory: The skill possesses high-privilege capabilities, including deleting dashboards, managing team memberships, and creating service account tokens, which increases the impact if an injection attack succeeds.
- Sanitization: There is no evidence of string sanitization or content validation performed on the ingested JSON data before it is presented to the agent for processing.
Audit Metadata