The Agent Skills Directory

[SAFE]: No security issues detected. The skill instructions and code examples are consistent with legitimate documentation for the Dask library.- [COMMAND_EXECUTION]: The skill utilizes dask.distributed.Client and functions like client.submit() or client.map() to execute Python tasks across worker nodes. This is the primary and intended purpose of the Dask library for parallel computation and does not involve unauthorized command execution.- [PROMPT_INJECTION]: The skill documents the ingestion of data from external files using read_csv, read_parquet, and read_text. While processing external data creates a potential surface for indirect prompt injection if the files contain malicious instructions, this is a standard data processing capability. * Ingestion points: Functions such as dd.read_csv(), dd.read_parquet(), and db.read_text() in SKILL.md and references/collections_guide.md. * Boundary markers: No specific delimiters or "ignore instructions" prompts are provided in the examples. * Capability inventory: Computational tasks are executed via client.submit(), client.map(), map_partitions(), and map_blocks(). * Sanitization: Standard data cleaning operations (e.g., dropna, fillna) are shown, though they are not designed to sanitize against adversarial instructions.

dask-parallel-computing