The Agent Skills Directory

[SAFE]: The skill documents the Modal cloud platform and its associated Python SDK. All identified behaviors, such as remote function execution, network operations, and file system management via Volumes, are consistent with the platform's intended purpose of orchestrating serverless cloud compute.
[COMMAND_EXECUTION]: The skill includes examples of executing subprocesses using subprocess.run and subprocess.Popen for tasks like starting web servers or launching distributed training scripts within containers. These are standard patterns for the documented tool.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the modal package and includes code for downloading machine learning models from established repositories like Hugging Face. These references target well-known, trusted services in the AI/ML ecosystem.
[DATA_EXFILTRATION]: While the skill interacts with sensitive authentication tokens and environment variables, it utilizes standard cloud security practices like secret management and encrypted environment injection. No unauthorized network operations or suspicious data transfer patterns were detected.
[PROMPT_INJECTION]: The skill processes external data in examples such as web scraping and LLM inference, which naturally creates a surface for indirect prompt injection. However, this is inherent to the domain of AI development tools, and the skill instructions do not contain any malicious injection or bypass patterns.

modal