The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The power-analysis-guide/SKILL.md file references gpower.hhu.de, which is the official website for the G*Power statistical software developed by Heinrich Heine University Düsseldorf. This is a well-known, trusted academic resource for sample size calculation.
[INDIRECT_PROMPT_INJECTION]: The guides contain code for processing external research datasets, which represents a standard indirect prompt injection surface for analytical agents.
Ingestion points: Functions across multiple files designed to take data arrays or dataframes as input, such as detect_univariate_outliers in data-anomaly-detection/SKILL.md and bayesian_regression in bayesian-statistics-guide/SKILL.md.
Boundary markers: None present; the reference code demonstrates direct data processing.
Capability inventory: Includes file system writing for logging artifacts (mlflow.log_artifact), data visualization (matplotlib), and complex statistical modeling.
Sanitization: None present; the code focuses on demonstrating statistical methodologies.
[COMMAND_EXECUTION]: The ml-experiment-tracker/SKILL.md guide demonstrates the use of mlflow to log experiment artifacts, including a step that writes a temporary file to /tmp/run_config.json. This is a standard and expected operation for experiment tracking tools.

statistics-skills