The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes the implementation of interceptors which, by design, ingest and process untrusted external data from the request lifecycle.
Ingestion points: Interceptor methods like onRequestStart and onRequestEnd in SKILL.md receive an event structure containing potentially untrusted data from CGI variables, request headers, and response buffers.
Boundary markers: No specific delimiters or safety warnings for handling untrusted data are provided in the template code, as the skill focuses on architectural implementation rather than data validation.
Capability inventory: The demonstrated interceptors have the capability to log data (log.info), modify request routing (event.overridePath), and manipulate the final response buffer (event.output).
Sanitization: The examples do not include explicit sanitization of the event payload, which is typical for architectural guides.
[DYNAMIC_EXECUTION]: The documentation describes mechanisms for dynamic registration and loading of classes.
Evidence: The skill demonstrates using context.loadClass() within the InterceptorService API and dynamic path interpolation in ModuleConfig.bx (#moduleRecord.invocationPath#.interceptors.RequestLogger) to register interceptors. These are standard features for the BoxLang modular system.

boxlang-core-dev-interceptors