The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large-scale external datasets from various formats such as CSV, Parquet, JSON, and text logs. This capability creates a potential surface for indirect prompt injection if the processed data is later used to influence agent behavior without proper sanitization.
Ingestion points: Data ingestion occurs via standard Dask functions like dd.read_csv, dd.read_parquet, da.from_zarr, and db.read_text as shown in SKILL.md and the reference files.
Boundary markers: The provided documentation does not include explicit instructions for using delimiters or boundary markers to separate external data from agent instructions.
Capability inventory: The skill enables parallel execution of arbitrary Python functions across partitions using map_partitions, map_blocks, and client.submit, as described in references/dataframes.md and references/futures.md.
Sanitization: There are no mentions of sanitization or validation techniques to filter potentially malicious instructions embedded within the data files.

dask