spark-engineer

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

• Third-party content exposure detected (medium risk: 0.65). The workflow ingests runtime data from external sources (e.g., spark.read.parquet("s3://bucket/events/") and streaming sources like Kafka/file paths), so any outsider-authored free text contained in those datasets can be read as LLM-readable strings and fed into the agent context.