The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/imiva_mcp.py executes the command npx -y @infimind/ecom-content-cli@latest. This pattern downloads and runs code from the NPM registry every time the skill is invoked, which introduces a dependency on the integrity of the remote registry and the package author.
[COMMAND_EXECUTION]: The helper script scripts/imiva_mcp.py uses subprocess.Popen to launch the external MCP client. While it uses a fixed package name, the reliance on shell-level execution to bridge functionality increases the attack surface.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data through the following evidence chain:
Ingestion points: User-provided inputs such as productImages and extraDescription are passed to the create_detail_page_task tool.
Boundary markers: None identified; the instructions do not specify delimiters to separate user data from agent instructions.
Capability inventory: The skill has the ability to execute shell commands via npx and perform network operations through the IMIVA API.
Sanitization: The Python helper script passes arguments to the MCP tool via JSON without evidence of input validation or filtering for malicious instructions embedded in the image metadata or description text.

ai-ecommerce-expert-ecommerce-detail-page