The Agent Skills Directory

[DYNAMIC_EXECUTION]: The load_model function in scripts/export_model.py uses pickle.load() to deserialize geological models from files with a .pkl extension. The pickle module is known to be insecure against erroneous or maliciously constructed data, as it can execute arbitrary code during the loading process. This pose a risk if the AI agent is directed to process a model file provided by an untrusted source.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data from CSV files containing surface points and orientations.
Ingestion points: External CSV data is loaded in scripts/validate_input.py and via code examples in SKILL.md using pandas.read_csv.
Boundary markers: None identified in the prompt templates or processing scripts.
Capability inventory: The skill includes capabilities to write files to the disk (np.save, df.to_csv, and open().write()) and perform complex 3D computations in scripts/export_model.py.
Sanitization: While scripts/validate_input.py performs basic numeric and column validation, it lacks comprehensive sanitization to prevent data-driven manipulation of the agent's logic.

gempy