The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes a code recipe for 'Production Caching' that utilizes joblib.load to deserialize Python objects from a local file. Because joblib (and the underlying pickle module) can execute arbitrary code during deserialization, loading files from untrusted sources presents a risk of code execution.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a surface for indirect prompt injection by ingesting machine learning datasets for feature attribution analysis.
Ingestion points: External data is passed to explainers in SKILL.md (e.g., explainer(X_test)).
Boundary markers: No explicit delimiters or warnings are provided to the agent to ignore instructions embedded within the data.
Capability inventory: The skill uses matplotlib for visualization and mlflow for experiment tracking.
Sanitization: No data sanitization or validation logic is present in the provided snippets.
[EXTERNAL_DOWNLOADS]: The skill references and recommends the installation of several well-known and reputable Python packages, including shap, matplotlib, xgboost, and mlflow. It also includes code to fetch standard example datasets from the official SHAP repository.

shap-model-explainability