smoke-test-ml-pipeline

Pass

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed to generate a pytest file for machine learning pipeline validation. It correctly utilizes standard Python libraries such as pytest and scikit-learn, and integrates with the vendor's own skrub package.
  • [PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by instructing the agent to read external files (experiment journals and scripts) and interpolate their contents into generated Python code. This represents a vulnerability surface rather than an active exploit.
  • Ingestion points: The skill reads journal/NN_*.md, experiments/NN_*.py, and src/<pkg>/data.py to derive test parameters and loader signatures.
  • Boundary markers: Absent; there are no instructions to use specific delimiters or ignore potentially malicious content within the analyzed files.
  • Capability inventory: The skill allows the agent to write a new Python test file to the tests/smoke/ directory.
  • Sanitization: Absent; values such as CV_MAE_MEAN and file paths are extracted and hardcoded into the output script without explicit validation or escaping.
Audit Metadata
Risk Level
SAFE
Analyzed
May 18, 2026, 03:59 PM
Security Audit — agent-trust-hub — smoke-test-ml-pipeline