ozon-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the
ozon-mcp-rupackage from the author's GitHub repository (github.com/ilyautov/ozon-mcp-ru) using theuvxtool. This is a standard deployment method for MCP servers. - [COMMAND_EXECUTION]: The skill configures the agent to execute the
uvxcommand to run the MCP server, passing API credentials as environment variables. - [INDIRECT_PROMPT_INJECTION]: The skill processes data retrieved from the Ozon API, which presents a potential surface for indirect prompt injection if the API data contains instructions meant for the AI agent.
- Ingestion points: Data is fetched from external Ozon API endpoints via the
ozon_call_methodtool. - Boundary markers: The instructions explicitly require the agent to show the user the request payload and wait for confirmation before performing any write operations.
- Capability inventory: The server has the capability to perform network requests (read/write) to the Ozon API.
- Sanitization: The skill implements a human-in-the-loop requirement where the user must approve all destructive or state-changing API calls.
Audit Metadata