The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process messages from Azure Service Bus queues and topics, which constitutes a potential surface for indirect prompt injection if external messages contain adversarial instructions designed to influence the agent.
Ingestion points: Messages are received via the receive_messages method demonstrated in SKILL.md and various processing patterns in references/patterns.md and references/dead-letter.md.
Boundary markers: The skill documentation explicitly recommends redacting payloads, using peek-lock modes for inspection, and implementing idempotency to mitigate risks associated with processing untrusted data.
Capability inventory: The bundled scripts/setup_servicebus.py utility can create, modify, and delete messaging entities such as queues, topics, and subscriptions.
Sanitization: The patterns reference provides guidance on validating message body types and handling correlation IDs to ensure data integrity.
[EXTERNAL_DOWNLOADS]: The skill instructions specify the installation of standard, well-known libraries (azure-servicebus, azure-identity, aiohttp) from the official Python Package Index (PyPI). Additionally, the skill references administrative support files sourced from Microsoft's public GitHub repository.
[COMMAND_EXECUTION]: The skill includes scripts/setup_servicebus.py, a management tool that executes commands to interact with cloud infrastructure. This script implements safety features, such as requiring an explicit --confirm-delete flag for destructive operations like deleting queues or topics.

azure-servicebus-py