skills/grafana/skills/ml-ai/Gen Agent Trust Hub

ml-ai

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous example curl commands intended for users to interact with Grafana Cloud's administrative APIs. These commands demonstrate how to create forecast jobs, verify metric production, and trigger Sift investigations. These are standard operational examples for the Grafana platform.
  • [INDIRECT_PROMPT_INJECTION]: The Grafana Assistant described in the skill processes natural language input to generate PromQL, LogQL, and TraceQL queries. This creates a surface where untrusted user input is translated into executable database queries.
  • Ingestion points: Natural language descriptions provided to the Grafana Assistant in references/llm-and-graph.md.
  • Boundary markers: Not explicitly defined in the provided markdown, relying on Grafana's internal implementation.
  • Capability inventory: The assistant can generate and explain queries, build dashboards, and correlate cross-signal telemetry (metrics, logs, traces).
  • Sanitization: The skill does not describe specific sanitization logic, as it documentation for an existing platform feature.
  • [EXTERNAL_DOWNLOADS]: The skill makes calls to grafana.net and grafana.com domains. These are official vendor resources for Grafana Cloud and are used here for legitimate API interactions and documentation links.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:02 AM