compute-mamba-ratio

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely technical and provides configuration guidance for a specific machine learning serving framework (SGLang).
  • [COMMAND_EXECUTION]: The Python code snippets included in the skill are for basic arithmetic calculations related to memory pools. They do not execute shell commands, access the file system, or perform network requests.
  • [PROMPT_INJECTION]: The content is strictly instructional regarding model parameters and does not contain any patterns intended to bypass AI safety filters or override system behavior.
  • [DATA_EXFILTRATION]: No network activity or sensitive data access was identified. The skill works with user-provided performance metrics and configuration flags.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 10:46 PM
Security Audit — agent-trust-hub — compute-mamba-ratio