The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for reading and processing data from external sources, creating a surface for potential indirect prompt injection attacks.
Ingestion points: The PartitionedReader class and the Iceberg metadata lookups in SKILL.md consume data from external Spark tables.
Boundary markers: The provided code templates do not include delimiters or instructions to treat ingested data as untrusted content.
Capability inventory: The skill utilizes powerful data transformation and write capabilities (insertInto, merge), which could be misused if instructions in the source data were interpreted as commands.
Sanitization: No content-level sanitization, escaping, or validation logic is present in the ETL class scaffolds.

pyspark-etl