The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for ingesting data from external Parquet files, which represents an inherent attack surface for indirect prompt injection if the source data is untrusted.\n
Ingestion points: The ParquetReader class is used to load data from external table_path locations.\n
Boundary markers: The skill relies on the structured nature of Parquet files and does not define explicit delimiters to separate data from instructions during processing.\n
Capability inventory: The skill is scoped to use tools like Bash and WebFetch for file operations and data fetching.\n
Sanitization: There is no specific mention of sanitizing or validating the content of the Parquet files before processing.

rerun-parquet