umap-learn
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructional content and code examples for the UMAP-learn library, a well-established tool for dimensionality reduction.
- [EXTERNAL_DOWNLOADS]: The skill recommends installing 'umap-learn' and related packages (scikit-learn, hdbscan, tensorflow) from official registries. These are well-known libraries in the data science community.
- [DYNAMIC_EXECUTION]: The skill mentions the use of Numba for compiling custom distance metrics. This is a standard optimization technique for the UMAP library and does not involve untrusted or remote code sources.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for data ingestion via data processing methods like 'fit' and 'transform' (SKILL.md, references/api_reference.md). While no explicit boundary markers are used for the input data, the skill advises using 'StandardScaler' for preprocessing. The available capabilities are restricted to mathematical operations and data clustering (hdbscan), presenting no risk of instruction override.
Audit Metadata