The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary installation method utilizes uvx to fetch and execute code directly from a remote, unverified GitHub repository (github.com/mpeirone/zabbix-mcp-server). This pattern executes arbitrary remote code on the user's machine during setup.
[EXTERNAL_DOWNLOADS]: In addition to the uvx execution, the documentation provides instructions to git clone the repository for local installation and Docker deployment, which involves downloading unverified scripts and configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge to Zabbix APIs, ingesting potentially untrusted data such as trigger descriptions, item values, and host metadata. This data is placed directly into the agent's context without documented boundary markers or sanitization, creating a surface for indirect prompt injection attacks.
Ingestion points: Monitoring data retrieved via the zabbix_api tool (e.g., host names, trigger descriptions, problem logs).
Boundary markers: None specified in the instructions to prevent the agent from obeying instructions embedded in monitoring data.
Capability inventory: The zabbix_api tool has high-privilege capabilities, including the ability to create, update, and delete Zabbix resources (unless restricted by environment variables).
Sanitization: No evidence of sanitization or filtering of API responses before they are processed by the AI agent.

zabbix-mcp-server