The Agent Skills Directory

[External Downloads] (LOW): The skill documentation includes functions for downloading datasets from remote repositories, such as download_all_regression and load_forecasting which fetch data from domains like forecastingdata.org and timeseriesclassification.com. These are standard features but represent unverified network connections.
[Prompt Injection] (LOW): The library exposes several data ingestion surfaces for formats like .ts, .arff, .tsv, and .csv (e.g., in references/datasets_benchmarking.md). This allows for potential indirect prompt injection if an agent uses these tools to process untrusted data files, as the results could influence downstream agent reasoning. Evidence: Ingestion points in datasets_benchmarking.md; no boundary markers described; capabilities include file processing and deep learning execution.
[Dynamic Execution] (INFO): The library documentation mentions adapters for external packages (e.g., PyODAdapter, SklearnClassifierWrapper) which wrap and execute logic from other libraries. This is a standard architectural pattern for integration in the scientific Python ecosystem.

aeon