ml-research-engineer-safeguards
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown documentation. There are no scripts, automated commands, or executable code included in the repository.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze safety datasets that include adversarial slices and indirect prompt injections. While this involves processing untrusted data, the instructions are focused on research analysis, labeling, and metric calculation rather than direct execution of the data content.
- [DATA_EXPOSURE_EXFILTRATION]: Guidelines include explicit constraints for PII (Personally Identifiable Information) handling and data redaction in training pipelines and dataset curation, which aligns with security best practices for research environments.
Audit Metadata