The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements model loading using joblib.load(). This function relies on Python's pickle module for deserialization, which is inherently insecure and can lead to arbitrary code execution if a malicious model file is processed. This is a common and accepted risk in machine learning workflows, but requires stringent access controls on model artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill defines API endpoints (/predict, /predict/batch) that ingest untrusted user data in the form of feature vectors. While the provided code utilizes Pydantic validation to enforce data types, it establishes a surface area where adversarial data could potentially influence the model's behavior.
Ingestion points: PredictionRequest and BatchPredictionRequest models in references/fastapi-production-server.md receive external feature vectors.
Boundary markers: The skill demonstrates the use of Pydantic schemas and Field validation to delimit and verify input data.
Capability inventory: The skill scripts utilize joblib for file loading, uvicorn for network serving, and include commands for Docker image building and Kubernetes deployment.
Sanitization: Pydantic validator decorators are used to ensure that input features are finite numbers, providing a baseline layer of input sanitization.

model-deployment