together-sandboxes

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install official Together AI SDKs (together for Python and together-ai for Node.js) to interact with the API. These are legitimate resources provided by the vendor.
  • [COMMAND_EXECUTION]: The skill's primary function is to execute Python code in a remote sandboxed environment. This includes support for runtime package installation via !pip install and shell commands within the sandbox as part of the intended code interpreter functionality.
  • [DATA_EXFILTRATION]: The skill uses the Together AI API (api.together.ai) to send code for execution. It correctly handles authentication using an environment variable (TOGETHER_API_KEY) rather than hardcoded credentials.
  • [INDIRECT_PROMPT_INJECTION]: As a code execution engine, the skill is designed to process and run agent-generated code. While this represents a capability that could be misused if the agent is compromised, the execution occurs within a managed remote sandbox, isolating the host environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 02:27 AM
Security Audit — agent-trust-hub — together-sandboxes