The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install umap-learn and its dependencies (such as hdbscan and tensorflow) from the official PyPI registry. These are widely recognized and trusted libraries within the data science community.
[DYNAMIC_EXECUTION]: The skill demonstrates the use of Numba's njit decorator to compile custom distance functions into machine code. This is a standard performance optimization used by the library for mathematical computations and does not involve untrusted code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external datasets, which theoretically creates an ingestion surface for indirect prompt injection. However, since the UMAP algorithm transforms data into numeric embeddings for visualization or clustering, there is no realistic risk of malicious instructions persisting or affecting the agent's behavior.
Ingestion points: Input data processed by the fit, fit_transform, and transform methods in SKILL.md.
Boundary markers: None present in the code examples.
Capability inventory: The skill utilizes Numba for JIT compilation and TensorFlow for neural network-based embeddings.
Sanitization: The skill relies on standard scikit-learn validation to ensure all input values are finite.

umap-learn