The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents features for ingesting data from external sources and local files, representing a standard attack surface for i18n tools.
Ingestion points: The file() function allows reading local files, and fetch() allows retrieving remote content for dictionaries.
Boundary markers: Data is structured within specific Intlayer content nodes (e.g., translation, enumeration).
Capability inventory: Supports filesystem access via fs and network operations.
Sanitization: The documentation encourages the use of the tagfilter option in Markdown and HTML rendering to mitigate risks from dangerous HTML tags.
[DYNAMIC_EXECUTION]: The library supports executing synchronous and asynchronous functions to generate dynamic content.
Evidence: Documentation in references/concept_content_function-fetching.md details the execution of content functions. It includes a security warning that this feature is disabled for remote content declarations to prevent unauthorized code execution.

intlayer-content