The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents the Smarty_RCMS environment, which is a server-side templating engine used for dynamic logic execution. The documentation in references/syntax.md highlights that the environment is restricted by a security sandbox that limits available functions and modifiers to a predefined list, preventing arbitrary PHP execution.
[DATA_EXFILTRATION]: The skill provides documentation for plugins such as {api}, {get_file}, and {sendmail} that allow the platform to read local data and transmit it to external services. However, it also emphasizes using the {secret} plugin for secure credential management, ensuring that API keys and tokens are not hardcoded or exposed in the templates.
[INDIRECT_PROMPT_INJECTION]: The skill describes the {ai_completion} plugin, which facilitates feeding data from API responses or files into LLMs, creating a surface for indirect injection.
Ingestion points: External data enters through {api} and {read_file} plugins as described in the API and file plugin references.
Boundary markers: The documentation recommends using escape and truncate modifiers when processing data.
Capability inventory: The environment supports network operations ({api}), file writes ({write_file}), and external notifications ({slack_post_message}).
Sanitization: The platform includes specialized sanitization modifiers such as escape and rcms_strip_tags to clean untrusted content.
[COMMAND_EXECUTION]: The skill documents high-level utility plugins like {generate_pdf}, {zip}, and {unzip}. These perform system-level tasks but are implemented as standard, non-interactive platform features that run asynchronously via Google Cloud Pub/Sub to maintain system integrity.

kuroco-server-processing