The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of data from external sources, creating a potential attack surface for indirect prompt injection.
Ingestion points: The skill instructions and PySpark reference file describe reading data from the Files/ directory and external cloud storage shortcuts, including Amazon S3, Google Cloud Storage, and Azure Data Lake Storage Gen2.
Boundary markers: There are no instructions or delimiters provided to isolate the agent from potentially malicious instructions embedded within the ingested data.
Capability inventory: The skill provides patterns for writing to managed tables, executing Spark SQL, and running Data Factory pipelines, which are significant capabilities if triggered by untrusted input.
Sanitization: The provided PySpark examples lack mechanisms for sanitizing or validating external data content before it is processed or used in query generation.
[DYNAMIC_EXECUTION]: The reference files include code patterns that use dynamic SQL execution.
Evidence: In references/pyspark.md, patterns such as the Incremental Load Pattern and SCD Type 2 Pattern utilize spark.sql() to execute queries constructed via string interpolation.

fabric-lakehouse