Skills
skills/orchestra-research/ai-research-skills/modal-serverless-gpu/Gen Agent Trust Hub

modal-serverless-gpu

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION] (LOW): The skill includes code templates that use subprocess.run(['nvidia-smi']). This is a legitimate operation to check GPU availability and drivers within a serverless container.
  • [EXTERNAL_DOWNLOADS] (LOW): The skill guides the user to install the modal CLI and specifies patterns for downloading machine learning libraries (transformers, torch) and container images from remote registries during environment setup.
  • [PROMPT_INJECTION] (LOW): The inference examples demonstrate an indirect prompt injection surface where untrusted data is processed by a model.
  • Ingestion points: The prompt parameter in the TextGenerator.generate method in SKILL.md.
  • Boundary markers: No delimiters or safety instructions are used to wrap the input prompt.
  • Capability inventory: The skill utilizes subprocess.run (different context), modal.Volume for persistent storage, and modal.Secret for credential management.
  • Sanitization: The example does not implement any validation or sanitization of the input text before passing it to the inference pipeline.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:06 PM