The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of content definitions that ingest data from external and internal sources, creating a potential attack surface for indirect prompt injection.
Ingestion points: Data is ingested via api field types (external API URLs), relation field types (references to other content records), and wysiwyg or html field types (direct user/editor input).
Boundary markers: The instructions do not specify boundary markers or "ignore instructions" prompts for handling data returned from external APIs or relations.
Capability inventory: Content structures can perform network operations via the api field and execute logic via Smarty templates.
Sanitization: The skill explicitly documents the allow_all_tags parameter for wysiwyg and html fields, which bypasses the default server-side sanitization that normally removes potentially dangerous tags like <script>.
[DYNAMIC_EXECUTION]: The skill provides instructions for configuring dynamic logic execution within content structures.
Template Evaluation: The api field type (ext_type 32) supports the use_smarty parameter, which enables the evaluation of configured URLs as Smarty templates at runtime.
JavaScript Hooks: The configuration allows specifying JavaScript hook names (on_select_hook, on_save_hook) that are executed during user interactions with the API picker.
Custom Processing: The skill mentions endpoint post-processing using api_method and append to dynamically manipulate content responses at runtime.

kuroco-content-structure