The Agent Skills Directory

[SAFE]: No malicious patterns or obfuscation detected. The skill serves as a template and instructional guide for machine learning evaluation.
[EXTERNAL_DOWNLOADS]: Python code snippets reference standard, well-known machine learning and data science libraries, including nltk, rouge-score, bert-score, transformers, detoxify, scikit-learn, scipy, and numpy.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths, or exfiltration patterns were discovered. The code utilizes standard API patterns for model interaction.
[PROMPT_INJECTION]: The skill defines a surface for processing untrusted data (evaluation test cases). While it lacks explicit boundary markers for these inputs in the provided snippets, this is standard for evaluation frameworks and no malicious override instructions were found.

llm-evaluation