zabbix-metrics-history
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a Zabbix monitoring server and processes it to provide answers to users, creating an attack surface where malicious data in monitoring items could attempt to influence agent behavior.
- Ingestion points: Data returned by the
zabbix_apitool (item.get,history.get,trend.get). - Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore instructions embedded within the retrieved monitoring data.
- Capability inventory: The skill uses
zabbix_apifor data retrieval. The documentation claims these tools are restricted to read-only operations via a destructive-method deny-list. - Sanitization: No evidence of sanitization or filtering of Zabbix item values or metadata before interpolation into the prompt.
- [EXTERNAL_DOWNLOADS]: The skill documentation identifies its underlying MCP server as a third-party component sourced from a repository (
mpeirone/zabbix-mcp-server) that is not within the trusted organization list. - [CREDENTIALS_UNSAFE]: The skill requires a
ZABBIX_TOKENfor operation. It correctly follows security best practices by requesting this via an environment variable (env: ["ZABBIX_TOKEN"]) rather than hardcoding credentials.
Audit Metadata