The Agent Skills Directory

[SAFE]: The skill describes a durable topic queue worker used to manage message delivery, retries, and dead-lettered messages. It defines standard functional triggers and configuration parameters.
[DATA_EXPOSURE]: Configuration examples reference local development endpoints (e.g., redis://localhost:6379, amqp://localhost:5672) and local file paths (./data/queue) for storage, which are standard for infrastructure components and do not expose sensitive information.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data by publishing JSON payloads to topics that are then consumed by subscriber functions.
Ingestion points: Payloads are ingested via the iii::durable::publish function defined in SKILL.md.
Boundary markers: None specified in the documentation for payload encapsulation.
Capability inventory: The skill triggers functions defined by function_id (e.g., 'orders::process') with no direct system capabilities exposed in this file.
Sanitization: No explicit sanitization or validation of the JSON payload is mentioned within the worker configuration.

queue