modal-serverless-gpu
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides code examples using
subprocess.run()for routine administrative tasks such as hardware status checks vianvidia-smiand package listing withpip list. These operations are standard for managing machine learning environments. - [EXTERNAL_DOWNLOADS]: Documentation includes instructions for downloading the
modallibrary and various ML packages (e.g.,transformers,torch) from official registries and trusted GitHub repositories. These downloads are necessary for the skill's stated purpose. - [INDIRECT_PROMPT_INJECTION]: The skill guides the user in creating web-facing endpoints that process user-provided strings for model inference. This represents a standard ingestion point for untrusted data inherent to machine learning deployment services.
- [DYNAMIC_EXECUTION]: The documentation demonstrates the use of the
modal.SandboxAPI to execute code in isolated environments, which is a legitimate and standard feature of the serverless platform described.
Audit Metadata