The Agent Skills Directory

[Data Ingestion Surface]: The skill provides patterns for parsing event payloads from JSON strings (e.g., webhook inputs). While this is a core functionality for messaging services, processing external data is a common surface for indirect prompt injection if the resulting data is subsequently used to influence agent behavior without sanitization.
Ingestion points: The methods EventGridEvent.fromString(jsonPayload) and CloudEvent.fromString(cloudEventJson) in SKILL.md are used to ingest untrusted data from external sources.
Boundary markers: The provided code snippets do not explicitly show boundary markers or delimiters for the ingested content.
Capability inventory: The skill is designed to publish events via sendEvent and sendEvents using the Azure Event Grid client.
Sanitization: The skill uses strongly-typed Java objects (POJOs) for data mapping, which provides structural validation, though it does not include explicit natural language sanitization logic.
[Credential Management]: The skill promotes secure development practices by demonstrating how to use DefaultAzureCredential and environment variables for managing API keys and access tokens.
Evidence: Examples in both SKILL.md and references/examples.md favor System.getenv() and AzureIdentityEnvVars over hardcoded secrets.
[Trusted Dependencies]: The skill utilizes official Maven dependencies from a well-known service provider, which is standard for professional software development.
Evidence: The skill references com.azure:azure-messaging-eventgrid and com.azure:azure-identity, which are official packages maintained by the vendor.

azure-eventgrid-java