managed-model-endpoints

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines a registration contract for model services that involves the generation of bash scripts for starting and stopping Docker containers. These scripts are intended to be executed by a system daemon.
  • [SAFE]: The provided script templates adhere to security best practices, specifically instructing the use of scoped environment variables (DOCKER_CONFIG) and piped input for credentials (docker login --password-stdin) to ensure secrets do not leak into process logs or the shell environment.
  • [SAFE]: The skill enforces network isolation by requiring that local containers bind strictly to the loopback interface (127.0.0.1), preventing unauthorized external access to the model endpoints.
  • [SAFE]: The documentation includes explicit warnings against using world-writable permissions (777) and correctly identifies the risks associated with multi-user environments.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 01:21 AM
Security Audit — agent-trust-hub — managed-model-endpoints