mamba-architecture
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No security issues detected. The skill provides legitimate documentation, configuration examples, and benchmarking workflows for the Mamba neural network architecture.
- [EXTERNAL_DOWNLOADS]: The skill refers to official and well-known Python packages (mamba-ssm, causal-conv1d, torch, transformers) for installation via standard package managers. These resources are hosted on public, established registries.
- [COMMAND_EXECUTION]: Includes standard shell commands for model benchmarking and distributed training (e.g., torchrun). These are typical for machine learning development environments.
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates text generation workflows where the model processes user-provided prompts. While this constitutes an attack surface inherent to language models, the skill uses these for their intended purpose without introducing specific vulnerabilities. 1. Ingestion points: User-supplied prompt input in the HuggingFace loading workflow (SKILL.md). 2. Boundary markers: None explicitly defined in the provided code snippets. 3. Capability inventory: Text generation via the model.generate method (SKILL.md). 4. Sanitization: None performed on the input prompt within the example snippets.
Audit Metadata