dask
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes workflows where the agent reads and processes external data from various file formats and remote storage (e.g., S3). This introduces an indirect prompt injection surface as the agent may process untrusted content that could contain embedded instructions intended to manipulate its logic.
- Ingestion points: The documentation (in
SKILL.md,references/dataframes.md, andreferences/bags.md) provides examples of reading data usingdd.read_csv(),db.read_text(),dd.read_parquet(), and glob patterns, which bring external content into the agent's processing context. - Boundary markers: The skill does not recommend or implement specific boundary markers or instructions to treat ingested data as non-executable text to prevent the agent from following instructions found within the data.
- Capability inventory: The skill utilizes Dask's powerful execution model, including functional transformations (
map,filter,apply) and distributed task submission (client.submit,client.map), which provide the capabilities necessary to execute complex operations influenced by the ingested data. - Sanitization: There is no mention of data sanitization, schema validation, or content filtering to mitigate risks from untrusted external inputs.
Audit Metadata