The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or exfiltration attempts were detected. The skill provides standard documentation and code examples for data science tasks using well-known, reputable libraries.
[REMOTE_CODE_EXECUTION]: The skill uses standard package management commands (uv pip install) to install the 'umap-learn' package. These are standard practices in a development environment and target official package registries.
[DYNAMIC_EXECUTION]: The skill mentions and provides examples for using Numba-compiled custom metrics and TensorFlow-based Parametric UMAP. These are standard high-performance computing and machine learning practices used for their intended purposes without signs of malicious intent or unsafe string evaluation.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external data (datasets for embedding), it treats them as numerical arrays for mathematical optimization. There is no evidence of the agent interpreting data content as natural language instructions, and the attack surface is limited to standard data science workflows.

alterlab-umap