The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data from Azure Event Hubs, which constitutes an indirect prompt injection surface if an agent processes this data without proper isolation.
Ingestion points: Data is consumed through the EventHubConsumerClient.subscribe method as demonstrated in the 'Receive Events' and 'Receive with Checkpointing' sections of SKILL.md.
Boundary markers: The provided implementation examples do not utilize delimiters or explicit instructions to the agent to treat event bodies as untrusted content.
Capability inventory: The skill enables reading from and writing to Azure Event Hubs and Azure Blob Storage. While the snippets primarily show logging event data, an agent implementing these methods could perform actions based on the content of the received events.
Sanitization: No input validation, escaping, or sanitization logic is shown for the event.body or event.properties before they are processed or logged.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to install official Node.js packages for Azure services.
Evidence: The documentation specifies installing @azure/event-hubs, @azure/identity, @azure/eventhubs-checkpointstore-blob, and @azure/storage-blob via npm. These are official, well-known packages maintained by Microsoft Azure.

azure-eventhub-ts