model-deployment
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The provided Python example uses joblib.load() to deserialize a model file. This method is inherently unsafe as it can execute arbitrary code if the model file is malicious or comes from an untrusted source. While common in ML workflows, it is a significant security risk compared to safer formats like ONNX.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of a REST API that ingests external data. 1. Ingestion points: The /predict endpoint in the generated app.py script. 2. Boundary markers: The implementation uses Pydantic schemas (PredictionRequest) to define expected input structures. 3. Capability inventory: The skill enables model loading and execution within a web server environment. 4. Sanitization: External content is sanitized via strict type validation (restricting features to a list of floats), reducing the risk of direct command or prompt injection, though the model remains vulnerable to adversarial data designed to bias inference results.
Audit Metadata