using-llm-specialist
Fail
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: CRITICAL
Full Analysis
- [PROMPT_INJECTION]: The automated scan identified the phrase 'ignore previous instructions' in
llm-safety-alignment.mdas a potential injection. However, this occurs within an educational context explaining how to defend against such attacks. The content does not contain instructions intended to override the agent's behavior. - [EXTERNAL_DOWNLOADS]: The skill references a wide range of external research papers, developer documentation, and open-source libraries. An automated scanner flagged
trulens.orgas malicious, but this is a well-established open-source tool for RAG evaluation (TruLens). All external references are appropriate for the skill's technical domain and are documented neutrally. - [COMMAND_EXECUTION]: The skill includes Python code snippets for fine-tuning and inference tasks. These snippets serve as templates and utilize standard, well-known libraries such as
transformers,peft, andtrl. No unauthorized or hidden command execution patterns were found.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata