The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes documentation in references/workflows.md demonstrating the use of the pickle module for serializing and deserializing models and inference data. While common in data science workflows for local persistence, using pickle.load on files from untrusted sources is insecure as it allows for arbitrary code execution during the deserialization process.
[INDIRECT_PROMPT_INJECTION]: The skill provides templates and scripts designed to ingest and process user-provided datasets, creating a surface for indirect prompt injection attacks. * Ingestion points: Data is loaded into predictors (X) and outcomes (y) in assets/linear_regression_template.py, assets/hierarchical_model_template.py, and referenced diagnostic scripts. * Boundary markers: The code and instructions lack explicit boundary markers or directives to treat external data as untrusted or to ignore instructions embedded within the data. * Capability inventory: The skill has the capability to write to the file system, including creating NetCDF files (idata.to_netcdf), CSV files (summary.to_csv), and saving various diagnostic plots as PNG images. * Sanitization: No input validation or sanitization routines are implemented to filter or neutralize potentially malicious content within the ingested datasets.

pymc