youtube-opus-gpt-benchmarks
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely instructional and does not contain any code, network requests, or sensitive file access patterns.
- [METADATA_POISONING]: The description mentions non-existent model versions (Claude 5.5, GPT-6.1). This is noted as thematic naming for benchmarks and does not appear to be a malicious attempt to bypass safety filters or deceive the agent about its capabilities.
Audit Metadata