industrial-ai-research
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill architecture is designed around structured research workflows with clear safety boundaries and non-negotiable standards for source verification. No malicious behaviors were detected.
- [EXTERNAL_DOWNLOADS]: The skill retrieves academic content from established services like arXiv and IEEE Xplore using standard WebSearch and WebFetch tools. These interactions are transparently documented as part of the intended research capability and target trusted academic infrastructure.
- [PROMPT_INJECTION]: The skill processes untrusted external data (academic papers) which presents a surface for indirect prompt injection. This risk is mitigated by the skill's architecture, which requires discrete metadata extraction, snippet verification, and a structured evidence-pack assembly phase rather than raw summarization. Ingestion points occur during source collection via WebFetch; capabilities are limited to standard read/web tools; sanitization is handled via verification and triage logic.
Audit Metadata