The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download a WordPress plugin from an untrusted GitHub repository release (https://github.com/msrbuilds/elementor-mcp/releases) and execute a Node.js package (@msrbuilds/emcp-proxy) via npx. Neither the author nor the repository belongs to the list of trusted organizations.
[REMOTE_CODE_EXECUTION]: The recommended proxy configuration uses npx @msrbuilds/emcp-proxy@latest, which downloads and executes remote code from a non-standard third-party source at runtime.
[DYNAMIC_EXECUTION]: The add-custom-code-snippet tool allows the AI agent to generate and inject arbitrary JavaScript and HTML code directly into the frontend of the WordPress site (head or body locations). This allows for dynamic script execution without human intervention.
[COMMAND_EXECUTION]: The installation steps provide shell commands for installing software via wp plugin install and running network proxies via npx and base64 for credential encoding.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Untrusted data from the WordPress site is ingested through tools like get-page-structure, get-element-settings, and search-stock-images (SKILL.md).
Boundary markers: The instructions lack delimiters or warnings to prevent the agent from following instructions embedded within the site's content.
Capability inventory: The skill possesses powerful capabilities including site-wide script injection (add-custom-code-snippet), media imports (sideload-image), and template creation (create-theme-template).
Sanitization: There is no evidence of sanitization or validation of the code or content generated by the AI before it is applied to the live site.

elementor-mcp-wordpress-builder