moysklad-mcp-install

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to clone a remote repository from GitHub (https://github.com/ilyautov/moysklad-mcp-ru) and execute scripts (install.py, serve.py) directly on the user's host machine when operating in a local CLI environment like Claude Code.
  • [DYNAMIC_EXECUTION]: The server script (serve.py) is designed to automatically create a Python virtual environment and install necessary dependencies upon its first execution, involving runtime package installation.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides tools to ingest data from the Moysklad JSON API (products, stocks, orders). This external data is processed by the agent, creating a surface for indirect prompt injection if the account content is controlled by a malicious actor.
  • Ingestion points: External business data enters the agent context via API tools such as ms_get_products and ms_get_stock defined in serve.py.
  • Boundary markers: The instructions do not specify the use of delimiters or specific tags to isolate API-returned data from the agent's core instructions.
  • Capability inventory: The agent possesses capabilities to execute shell commands, manage local files, and perform network requests to the Moysklad API.
  • Sanitization: The skill instructions do not outline specific validation or sanitization procedures for the data retrieved from the external API before it is presented to the LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 02:17 PM
Security Audit — agent-trust-hub — moysklad-mcp-install