mamba-architecture
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access behaviors were detected across the skill files.
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install standard research libraries (
mamba-ssm,causal-conv1d) from PyPI and fetches pretrained model weights from HuggingFace repositories (e.g.,state-spaces/mamba-2.8b). These are official sources for the Mamba architecture and represent routine development tasks. - [INDIRECT_PROMPT_INJECTION]: The workflow for text generation involves processing user prompts through the loaded models. This follows standard model interaction patterns and does not involve instructions to bypass safety guardrails or execute unauthorized actions.
Audit Metadata