The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for building applications that process untrusted user data via forms and APIs, creating a potential attack surface. 1. Ingestion points: Untrusted data enters through Resource::formSchema() and Model::dataSchema() as described in SKILL.md and references/model-and-resource.md. 2. Boundary markers: The skill mandates the use of the FormField facade and Model::validate() methods for all inputs to ensure consistency and isolation. 3. Capability inventory: The framework provides capabilities for file system writes via MediaFileManager, database persistence, and dynamic CRUD route generation. 4. Sanitization: Data is sanitized using the Field::format() method and processed through theme-specific renderers (Wonder/Bootstrap) to prevent raw HTML injection.
[COMMAND_EXECUTION]: The skill instructs the agent to use CLI tools for development and validation, including php forge update --local, php forge start, composer dumpautoload, and herd restart.
[DYNAMIC_EXECUTION]: The framework architecture relies on dynamic discovery and loading of PHP classes from multiple directory layers (core, modules, site, and custom) via ModelRegistry and ResourceRegistry.

wi-app