The Agent Skills Directory

[SAFE] (SAFE): No malicious patterns detected. The skill provides legitimate documentation and code examples for the Modal cloud platform.
[COMMAND_EXECUTION] (SAFE): Includes a snippet using subprocess.run(['nvidia-smi']) to retrieve GPU information. This is a standard diagnostic command and poses no risk in this context.
[EXTERNAL_DOWNLOADS] (SAFE): References installation of the modal library and various ML dependencies (transformers, torch) via standard package managers. These are intended for environment setup.
[PROMPT_INJECTION] (LOW): As a skill that demonstrates building LLM inference endpoints, it naturally has an indirect prompt injection surface (Category 8) where user-provided prompts are passed to models.
Ingestion points: predict(text: str) and generate(self, prompt: str) methods in the provided code snippets.
Boundary markers: Absent in snippets.
Capability inventory: Deployment of web endpoints and serverless functions.
Sanitization: Not shown in the basic examples provided. This is an inherent property of the task and is noted for awareness.

modal-serverless-gpu