mage-ai-pipelines

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Streaming Kafka messages are consumed by the required runtime workflow and passed into the transformer block (transform_events(messages: list, ...)), meaning outsider-authored free text from the Kafka topic can reach the agent via messages without selecting a specific item first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime dependencies that fetch and execute remote code — the Docker image "mageai/mageai:0.9.72" (pulled and run by Docker/Kubernetes) and Spark jars "s3a://jars/delta-core_2.12-2.4.0.jar" and "s3a://jars/hadoop-aws-3.3.4.jar" (downloaded and loaded into Spark) which meet the criteria of external content executed at runtime.