The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill suggests installing the judgy Python package (pip install judgy) to implement the Rogan-Gladen bias correction formula and confidence interval estimation. This is an external third-party dependency relevant to the skill's mathematical context.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting human-labeled data sets for evaluation purposes. • Ingestion points: The labeled_data structure referenced in Python code snippets within SKILL.md. • Boundary markers: No explicit prompt boundaries or delimiters are specified for the training or evaluation data. • Capability inventory: The skill is restricted to statistical analysis and data manipulation using numpy and scikit-learn; it does not contain network, file-system write, or subprocess execution capabilities. • Sanitization: The skill does not implement specific sanitization or filtering for the input data strings as it does not perform high-risk actions with the data.

validate-evaluator