real-time-streaming
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for processing unbounded data streams from external sources (Kafka, Flink), which creates a potential surface for indirect prompt injection if the ingested data contains malicious instructions.
- Ingestion points: Kafka topics (e.g.,
orders,clicks,customers) and Flink DataStreams described inSKILL.md. - Capability inventory: The skill demonstrates writing data to external sinks (JDBC, custom sinks) and configuring database connectors via Debezium and Kafka Connect.
- Boundary markers: The provided Java and Bash examples do not explicitly define delimiters or boundary markers to differentiate between data and potentially embedded instructions.
- Sanitization: The transformation and aggregation examples (e.g.,
CountAggregator,process(record)) do not show explicit validation or sanitization of the event content, which is standard for architectural examples but worth noting as a risk surface.
Audit Metadata