The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/imiva_mcp.py invokes npx -y @infimind/ecom-content-cli@latest, which automatically downloads and executes the latest version of the specified package from the NPM registry. This behavior allows for the execution of remote code at runtime without explicit version pinning or integrity checks.\n- [COMMAND_EXECUTION]: The skill uses the subprocess.Popen function in Python to call system binaries (specifically npx). While this is used to initiate the MCP communication bridge, it represents a direct interface for executing shell commands from the agent environment.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill is designed to handle sensitive credentials in the form of an MCP_TOKEN passed via environment variables. The helper script transmits this token to the external domain imiva.ecpro.com to facilitate authenticated API requests. While documented as the core functionality, this pattern involves the transit of user secrets to third-party infrastructure.\n- [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by ingesting and processing untrusted data from the user and potentially from image metadata.\n
Ingestion points: The prompt and images parameters in the create_smart_refine_task tool, handled by the scripts/imiva_mcp.py script.\n
Boundary markers: None; input strings and image paths are interpolated into tool calls without delimiters or instructions to ignore embedded commands.\n
Capability inventory: The skill possesses capabilities for remote code execution, command execution via subprocess, and network access to external endpoints.\n
Sanitization: No validation or sanitization logic is present to filter malicious instructions or malformed paths from the prompt and image arguments before they are sent to the external processor.

ai-ecommerce-expert-chinese-text-ad-poster