The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for ingesting user-submitted form data via the $request object in the PHP do_action method. This data is intended to be passed to external APIs or CRMs, creating a vulnerability surface where maliciously crafted form inputs could target downstream systems.
Ingestion points: Untrusted data enters the skill's context through the $request parameter of the do_action( array $request, Action_Handler $handler ) method in SKILL.md.
Boundary markers: The implementation examples do not use specific boundary delimiters for user-provided data, but the documentation provides structural guidance for field mapping.
Capability inventory: The PHP class implementation includes network operation capabilities through the call_remote_api helper method, which is intended to send form data to external endpoints.
Sanitization: The skill explicitly recommends security hygiene in its "Critical rules" section, advising the use of is_email(), absint(), and sanitize_text_field() to validate and clean data before it is processed or transmitted.

jfb-form-action