The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates for Azure Functions that ingest data from untrusted sources, creating a surface for indirect prompt injection attacks.
Ingestion points: The skill templates in references/detailed-guide.md process external data via HttpRequest (parameters and body), StorageBlob (blob names/paths), and QueueTrigger (message payloads) across .NET, Node.js, and Python examples.
Boundary markers: The provided code templates do not implement explicit delimiters or instruction boundary markers to separate untrusted input from the agent's logic.
Capability inventory: The skill templates utilize HttpClient for network operations and include Azure Storage bindings (QueueOutput, StorageBlob) for data persistence. Deployment instructions reference the az CLI for cloud resource configuration in references/detailed-guide.md.
Sanitization: While the code snippets are templates, the documentation in references/detailed-guide.md includes specific security guidance to "Validate input, ownership and idempotency before enqueuing" and to "Prevent SSRF with destination allowlists... Never trust arbitrary URLs."

azure-functions