The Agent Skills Directory

[SAFE]: The skill functions as an educational guide for the Dask library, providing standard Python code snippets for data analysis and distributed task scheduling. The included code examples follow established library patterns and do not contain malicious commands or unauthorized network operations.
[INDIRECT_PROMPT_INJECTION]: The skill documents workflows for ingesting and processing untrusted external data (such as CSV, JSON, and text logs) using Dask collections. This establishes an attack surface for indirect prompt injection where malicious instructions embedded in data files could influence the agent during processing.
Ingestion points: dd.read_csv, db.read_text, and da.from_zarr calls in SKILL.md.
Boundary markers: No explicit boundary markers or delimiters for untrusted data are specified in the instructional code blocks.
Capability inventory: The skill utilizes Python execution capabilities including compute(), map_partitions(), map_blocks(), and client.submit() for distributed processing.
Sanitization: The provided examples do not include input validation or sanitization of external data content.
Note: This risk is considered inherent to the primary purpose of a data processing library and does not represent a specific vulnerability in the skill's implementation.
[COMMAND_EXECUTION]: While Dask involves task execution across cores or clusters, the skill uses standard library methods like dask.distributed.Client for resource management. There are no instances of arbitrary shell command execution or privilege escalation attempts.

dask