dask
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill documentation describes patterns for ingesting and processing data from external, potentially untrusted sources (such as CSV, Parquet, JSON, and log files), which creates an indirect prompt injection vulnerability surface. \n
- Ingestion points: Data ingestion methods are documented in
SKILL.md(dd.read_csv),references/dataframes.md(dd.read_csv,dd.read_parquet), andreferences/bags.md(db.read_text,db.from_sequence). \n - Boundary markers: The instructions do not include the use of delimiters or guidance to ignore embedded instructions within data partitions. \n
- Capability inventory: The skill utilizes extensive capabilities, including file system access, network coordination via distributed schedulers, and arbitrary Python function execution through
client.submit,map_partitions, andmap_blocks. \n - Sanitization: There is no documentation recommending the validation, filtering, or sanitization of data content before processing.
Audit Metadata