The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs on building modules that ingest data from external sources (the Joomla database) and render it within templates. This creates a surface for indirect prompt injection if an attacker can manipulate database content (e.g., article titles, user names) to include instructions for an agent.
Ingestion points: src/Helper/ExampleHelper.php and src/Helper/ArticlesHelper.php fetch data from the #__articles, #__categories, and #__users tables.
Boundary markers: The instructions and examples lack specific boundary markers or instructions for the agent to ignore embedded commands within the rendered content.
Capability inventory: The skill utilizes database read operations (DatabaseInterface) and dynamic file inclusion (require in mod_ejemplo.php).
Sanitization: The skill correctly recommends and demonstrates the use of htmlspecialchars() and Joomla\CMS\HTML\HTMLHelper to sanitize output for HTML rendering, which mitigates XSS but not necessarily prompt injection.
[DYNAMIC_EXECUTION]: The entry point script (mod_ejemplo.php) uses dynamic file inclusion via require $path, where the path is resolved at runtime based on the layout parameter. While this is standard practice in Joomla's architecture, it involves executing code from a path computed from configuration data.

joomla-module-development