nemo-curator
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructions for data curation using the NeMo Curator toolkit. All cited dependencies (e.g., nemo-curator, cudf, rapids) and models (e.g., nvidia/quality-classifier-deberta, sentence-transformers/all-MiniLM-L6-v2) are standard tools from reputable sources.
- [SAFE]: The skill encourages security best practices by including modules for PII redaction and NSFW content detection as part of the data curation pipeline.
- [SAFE]: No suspicious command execution, persistence mechanisms, or unauthorized data access patterns were detected.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as web scrapes (Common Crawl) and S3 bucket contents. Ingestion points: DocumentDataset.read_parquet calls in SKILL.md. Boundary markers: None explicitly defined in example scripts. Capability inventory: File writing via dataset.to_parquet. Sanitization: Built-in PII redaction, quality filtering, and NSFW detection serve as mitigations.
Audit Metadata