The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches source code and installation scripts from the vendor's repository at github.com/ilyautov/marketplaces-mcp-ru using git clone or ZIP downloads via curl.
[COMMAND_EXECUTION]: Executes the install.py script to perform the local setup and serve.py to verify server readiness. It also uses standard CLI commands like uname, whoami, and python3 to detect the environment and verify requirements.
[PERSISTENCE]: Modifies agent configuration files (such as claude_desktop_config.json and opencode.json) to register the marketplace MCP servers, ensuring they remain active across sessions.
[PRIVILEGE_ESCALATION]: Recommends bypassing macOS security controls by using xattr -d com.apple.quarantine on downloaded scripts if they are blocked from executing by the operating system.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from various Marketplace APIs, which represents a potential attack surface if the API responses contain malicious instructions.
Ingestion points: Fetches product descriptions, stock levels, and sales data via Wildberries and Ozon API endpoints.
Boundary markers: The instructions do not specify explicit delimiters or warnings for the agent to ignore instructions embedded in the retrieved marketplace data.
Capability inventory: The agent has the ability to write to the local filesystem (cabinets.json), modify client configurations, and execute Python subprocesses.
Sanitization: No specific data validation or sanitization procedures are outlined in the skill instructions for handling the remote API payloads.

marketplace-mcp-install