Skills
Navigate
skills/zapabob/skills/modal-serverless-gpu/Gen Agent Trust Hub

modal-serverless-gpu

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The code examples for ML inference ingest user-supplied text prompts, creating a surface for indirect prompt injection. This behavior is inherent to the machine learning tasks the skill documents.
  • Ingestion points: TextGenerator.generate, predict, and batch_predict (SKILL.md); predict and monitored_inference (references/advanced-usage.md).
  • Boundary markers: None provided in the illustrative code examples.
  • Capability inventory: subprocess.run, modal.Volume, modal.Secret (SKILL.md); modal.Sandbox (references/advanced-usage.md).
  • Sanitization: No input sanitization or validation logic is present in the provided snippets.
  • [COMMAND_EXECUTION]: The skill instructs the user to run modal setup for authentication and demonstrates the use of subprocess.run to execute nvidia-smi and torch.distributed.launch for diagnostics and training orchestration within the remote environment.
  • [EXTERNAL_DOWNLOADS]: The skill documentation includes examples that fetch the transformers library directly from Hugging Face's official GitHub repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 12:08 PM